Corporate Spend Optimization and Mapping Model Architecture

Abstract

Various embodiments described herein relate to providing optimization related to enterprise performance management. In this regard, a request is received to obtain one or more insights regarding a formatted version of disparate data associated with one or more data sources. The request includes an insight descriptor that describes a goal for the one or more insights. In response to the request, aspects of the formatted version of the disparate data are associated to provide the one or more insights. The associated aspects are determined by the goal and a relationship between the aspects of the formatted version of the disparate data. Further, one or more actions are performed based on the one or more insights.

Description

[Technical Field]

[0001] (CROSS-REFERENCE TO RELATED APPLICATIONS) This application claims the benefit of U.S. Provisional Application No. 63 / 072,560, entitled "UNCLASSIFIED SPEND OPTIMIZATION," filed August 31, 2020, and U.S. Provisional Application No. 63 / 149,004, entitled "ENTERPRISE SPEND OPTIMIZATION AND MAPPING MODEL ARCHITECTURE," filed February 12, 2021, the contents of which are incorporated herein by reference in their entireties.

[0002] FIELD OF THE INVENTION The present disclosure relates generally to machine learning and, more particularly, to optimization related to enterprise performance management. Summary of the Invention

[0003] According to an embodiment of the present disclosure, a method is provided. The method enables a device having one or more processors and memory to receive a request to obtain one or more insights regarding a formatted version of disparate data associated with one or more data sources. The request includes an insight descriptor that describes a goal for the one or more insights. The method also enables the device to correlate aspects of the formatted version of the disparate data in response to the request to provide the one or more insights, the associated aspects being determined by a relationship between the goal and the aspects of the formatted version of the disparate data. The method also enables the device to perform one or more actions based on the one or more insights in response to the request.

[0004] According to another embodiment of the present disclosure, a system is provided. The system includes one or more processors, a memory, and one or more programs stored in the memory. The one or more programs include instructions configured to receive a request to obtain one or more insights about a formatted version of disparate data associated with one or more data sources. The request includes an insight descriptor that describes a goal for the one or more insights. The one or more programs also include instructions configured to correlate aspects of the formatted version of the disparate data to provide the one or more insights in response to the request, the associated aspects being determined by a relationship between the goal and the aspects of the formatted version of the disparate data. The one or more programs also include instructions configured to perform one or more actions based on the one or more insights in response to the request.

[0005] According to yet another aspect of the present disclosure, a non-transitory computer-readable storage medium is provided. The non-transitory computer-readable storage medium includes one or more programs executed by one or more processors of a device. The one or more programs include instructions that, when executed by the one or more processors, cause the device to receive a request to obtain one or more insights regarding formatted versions of disparate data associated with one or more data sources. The request includes an insight descriptor that describes a goal for the one or more insights. The one or more programs also include instructions that, when executed by the one or more processors, cause the device to correlate aspects of the formatted versions of the disparate data to provide the one or more insights in response to the request, the associated aspects being determined by a relationship between the goal and aspects of the formatted versions of the disparate data. The one or more programs also include instructions that, when executed by the one or more processors, cause the device to perform one or more actions based on the one or more insights in response to the request. [Background technology]

[0006] Traditionally, a large portion of the time (e.g., 60%-80% of the time) involved in data analysis and / or digital transformation of data involves cleaning and / or preparing the data for analysis. Additionally, limited time is traditionally spent modeling the data to, for example, provide insights related to the data. Thus, computing resources involved in data analysis and / or digital transformation of data are traditionally used in an inefficient manner. [Brief explanation of the drawings]

[0007] The description of the illustrated embodiments may be read in conjunction with the accompanying drawings. It will be understood that for simplicity and clarity of illustration, elements shown in the drawings have not necessarily been drawn to scale. For example, the dimensions of some of the elements are exaggerated relative to other elements. Embodiments incorporating the teachings of the present disclosure will be shown and described in connection with the figures presented herein. [Figure 1] 1 illustrates an exemplary networked computing system environment in accordance with one or more embodiments described herein. [Figure 2] 1 illustrates a schematic block diagram of an IoT platform framework for a networked computing system, according to one or more embodiments described herein. [Figure 3] 1 illustrates a system that provides an exemplary environment, according to one or more embodiments described herein. [Figure 4] 1 illustrates another system providing an exemplary environment, according to one or more embodiments described herein. [Figure 5] 1 illustrates an exemplary computing device in accordance with one or more embodiments described herein. [Figure 6] 1 illustrates a system for facilitating optimization related to enterprise performance management, according to one or more embodiments described herein. [Figure 7]1 illustrates a machine learning model according to one or more embodiments described herein. [Figure 8] 1 illustrates a system associated with an exemplary mapping model architecture, according to one or more embodiments described herein. [Figure 9] 1 illustrates a system associated with another example mapping model architecture, according to one or more embodiments described herein. [Figure 10] 1 illustrates a system associated with an exemplary transformer-based classification model, according to one or more embodiments described herein. [Figure 11] 1 illustrates a system associated with an exemplary neural network architecture, according to one or more embodiments described herein. [Figure 12] 1 illustrates a flow diagram for performing optimization related to enterprise performance management, according to one or more embodiments described herein. [Figure 13] 1 illustrates a flow diagram for performing optimization related to enterprise performance management, according to one or more embodiments described herein. [Figure 14] 1 illustrates a functional block diagram of a computer that may be configured to perform the techniques described in accordance with one or more embodiments described herein. [Figure 15] 1 illustrates an exemplary user interface according to one or more embodiments described herein. [Figure 16] 10 illustrates another exemplary user interface according to one or more embodiments described herein. [Figure 17] 10 illustrates yet another exemplary user interface according to one or more embodiments described herein. DETAILED DESCRIPTION OF THE INVENTION

[0008] Reference will now be made in detail to the embodiments, examples of which are illustrated in the accompanying drawings. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the various described embodiments. However, those skilled in the art will understand that the various described embodiments may be practiced without these specific details. In other instances, well-known methods, procedures, components, circuits, and networks have not been described in detail so as not to unnecessarily obscure aspects of the embodiments. The term "or" is used herein in both the alternative and conjunctive sense, unless otherwise indicated. The terms "illustrated," "example," and "exemplary" are used as examples without an indication of level of quality. Like numerals refer to like elements throughout.

[0009] The phrases "in an embodiment," "in one embodiment," "according to one embodiment," and similar phrases generally mean that the particular feature, structure, or characteristic that follows the phrase may be included in at least one embodiment of the present disclosure, and may be included in more than one embodiment of the present disclosure (importantly, such phrases do not necessarily refer to the same embodiment).

[0010] The word "exemplary" is used herein to mean "serving as an example, instance, or illustration." Any implementation described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other implementations.

[0011] When the specification states that a certain component or feature "can include," "may include," "could include," "should include," "would include," "preferably include," "possibly include," "typically include," "optionally include," "for example include," "often include," or "might include" (or other such language), or has a certain characteristic, the particular component or feature is not required to be included or to have that characteristic. Such components or features may be optionally included or excluded in some embodiments.

[0012] Generally, the present disclosure provides an "Internet of Things" or "IoT" platform for enterprise performance management that uses real-time models, near real-time models, and visual analytics to deliver rational, actionable recommendations for sustained peak performance of a business or organization. The IoT platform is a portable, scalable platform that can be deployed in any cloud or data center environment to provide a top-down view of the entire enterprise, displaying the status of processes, assets, people, and safety. Additionally, the IoT platform of the present disclosure supports end-to-end capabilities to run digital twins on process data and convert the output into actionable insights, as described in more detail below.

[0013] Traditionally, a significant portion of the time involved in data analysis and / or digital transformation of data (e.g., greater than 50% of the time, 60%-80% of the time, etc.) involves cleaning and / or preparing the data for analysis. Furthermore, traditionally, limited time is spent modeling the data to, for example, provide insights related to the data. Thus, computing resources involved in data analysis and / or digital transformation of data are traditionally used inefficiently.

[0014] As an example, enterprises often have procurement organizations that optimize spending (e.g., resource usage, asset usage, etc.) through various processes related to assets and / or services. However, it is generally difficult for procurement organizations to make spending decisions due to the scale (e.g., number of assets, number of parts, number of suppliers, etc.) and / or complexity (e.g., different geographic regions, different contracts, different vendors, etc.) of the enterprise's spending information. For example, procurement professionals generally do not have all the context available to make spending decisions, such as whether to negotiate a 60-day or a 90-day payment term in an asset and / or service contract. Furthermore, it is generally difficult for procurement professionals to determine where to prioritize their efforts to maximize value for the enterprise. For example, it is generally difficult for procurement professionals to determine whether to renegotiate asset and / or service contracts or instead consolidate assets and / or services. In this regard, traditional data analysis techniques generally result in inefficient use of computing resources, increased storage requirements, and / or increased error-related data. Furthermore, traditional data processing generally does not scale with the complexity of the data processing. It should also be appreciated that other technical challenges may exist with traditional data analysis and / or the digital transformation of traditional data.

[0015] Accordingly, to address these and / or other issues, examples of optimization related to enterprise performance management are provided. Various embodiments described herein relate to unclassified data optimization for an enterprise. For example, various embodiments described herein relate to unclassified spend optimization. Unclassified spend optimization includes, for example, unclassified spend optimization for assets, unclassified spend optimization for industrial plants, unclassified spend optimization for warehouses, unclassified spend optimization for buildings, unclassified spend optimization for an enterprise, and / or another type of unclassified spend optimization related to spending goals. Various embodiments described herein additionally or alternatively relate to unclassified asset optimization. Various embodiments described herein additionally or alternatively relate to optimization for supply chain analytics. For example, various embodiments described herein additionally or alternatively relate to optimization related to shipping terms. Various embodiments described herein additionally or alternatively relate to other types of optimization related to enterprise performance management. Enterprise performance management includes, for example, asset performance management, industrial plant performance management, warehouse performance management, building performance management, enterprise performance management, and / or performance management of another type of optimization goal. Various embodiments described herein additionally or alternatively provide a mapping model architecture related to formatting heterogeneous data associated with one or more data sources. Further, in various embodiments described herein, one or more characteristics associated with the format structure of the heterogeneous data are inferred to provide one or more mapping recommendations for the formatted version of the heterogeneous data. In one or more embodiments, the one or more mapping recommendations facilitate the transfer of data between a first data source and a second data source. In one or more embodiments, the one or more mapping recommendations facilitate one or more machine learning processes associated with the heterogeneous data. In one or more embodiments, the one or more mapping recommendations facilitate providing one or more insights associated with the heterogeneous data. In one or more embodiments, the one or more mapping recommendations facilitate the performance of one or more actions based on the heterogeneous data.

[0016] In various embodiments, optimization related to enterprise performance management provides insights (e.g., actionable insights) across enterprise domains with scalable data liquidity. For example, in various embodiments, data-driven opportunities are identified by using intelligent data processing to generate value for data in a reduced time (e.g., seconds, minutes, hours, days, or weeks) compared to traditional data processing systems. In various embodiments, a data liquidity layer is provided across the enterprise by using artificial intelligence to automate data integration to provide a network of knowledge that can be used for data analysis and / or digital transformation to create value for data. In various embodiments, multi-domain artificial intelligence offerings are provided and / or realized via one or more network or cloud computing environments.

[0017] In various embodiments, data from one or more data sources (e.g., relational data sources, data exchange data sources, comma-separated value data sources, and / or another type of data source) is ingested and data preparation and / or data blending is facilitated for the data. In various embodiments, one or more intelligent machine learning systems (e.g., one or more intelligent machine learning bots) map data from different sources into a common data format. In various embodiments, a mapping file is used to map each data field from the data collected from the sources to create a denormalized database. In various embodiments, data deduplication, rationalization, autofill, and / or anomaly detection are additionally or alternatively performed on the data to facilitate data mobility at scale. In various embodiments, enterprise semantics (e.g., industry semantics) are overlaid on the data to provide real-world meaning across enterprise systems and / or provide enterprise-wide applications. In various embodiments, an artificial intelligence recommendation engine provides persona-based recommendations for spend classification, product classification, product reclassification, payment term optimization, risk mitigation, alternative supplier identification, and / or other insights for enterprise optimization.

[0018] In various embodiments, data from one or more data sources is ingested, cleaned, and aggregated to provide aggregated data. Further, in various embodiments, one or more insights are determined from the aggregated data to provide cost savings and / or efficiency insights. In one or more embodiments, data is extracted from one or more data sources and unified into a single data lake. A data lake is a storage repository that stores data, for example, as raw data and / or in the data's original format. In one or more embodiments, the data lake is updated at one or more predetermined intervals to keep the data in the data lake current. According to one or more embodiments, data in the data lake is unified by recognizing different fields in the data lake as describing the same subject (e.g., vendor name, payment term, etc.) and / or by arranging all available terms (e.g., corresponding subjects) in the same format. In one or more embodiments, one or more operations are performed to complete data sources with incomplete field information (e.g., by recognizing the missing field as the same field in another data source with complete information and using that information to supply the missing information, etc.).

[0019] In one or more embodiments, data in the data lake is organized in an ontology structure. In one or more embodiments, the ontology structure enables understanding of the complex structures associated with the complex relationships between the disparate data in the data lake (e.g., "show all vendors in a particular geographic location whose products supplied in that geographic location depend on commodity X," "show all purchase orders where shipments were made Y days late," "show all industrial assets in an industrial plant where some degree of inefficiency exists for an interval of Z days," "show all work order requests in an industrial plant where maintenance delays have resulted in some degree of inefficiency," etc.). In one or more embodiments, based on the organized structure of the data lake, data sources are periodically compared to identify and provide one or more opportunities for cost savings and / or efficiency. For example, based on the organized structure of the data lake, it can be determined that payment terms for the same supplier are different in two different purchase orders and should be the same. In another example, based on the organized structure of the data lake, it can be determined that the price of a commodity is cheaper from a second supplier. In yet another example, it may be determined that the cost of a commodity is cheaper on the open market and therefore it would be more efficient to break or renegotiate the current contract for the commodity. In yet another example, part master data (e.g., a single source of parts) is created by ingesting data from multiple data sources to maintain different part numbers and / or provide uniform visibility across the enterprise. In yet another example, a unified procurement database is provided that relates data from multiple enterprise systems to facilitate metrics insights across disparate enterprise systems.

[0020] In one or more embodiments, unclassified data about an organization is collected, cleansed, and / or aggregated to facilitate delivery of one or more actions generated by one or more artificial intelligence (AI) models. According to various embodiments, one or more AI models are used to prioritize actions to be performed by a procurement organization, e.g., to maximize value for the procurement organization. According to various embodiments, data mapping of the unclassified data (e.g., unclassified data from multiple source systems) is performed to convert the unclassified data into an internal representation used by one or more AI models. According to various embodiments, one or more AI models are trained to determine one or more inferences and / or classifications for the unclassified data.

[0021] In one or more embodiments, deep learning (e.g., deep learning associated with one or more AI models) is performed to determine part commodity families for uncategorized purchase record data obtained from multiple data sources. According to one or more embodiments, the purchase record data includes, for example, purchase order data, vendor data (e.g., customer-vendor data), invoice data, and / or other data. In embodiments, the uncategorized purchase record data is obtained from multiple external data sources. Additionally or alternatively, in another embodiment, the uncategorized purchase record data is obtained from a cloud database. Further, in one or more embodiments, total spend for part commodity families is aggregated to provide categorized purchase record data. In one or more embodiments, one or more actions are performed based on the categorized purchase record data.

[0022] In one or more embodiments, field mapping is used to migrate data between databases, data models, and / or systems. In one or more embodiments, field mapping uses entity relationships to facilitate data migration between databases, data models, and / or systems. In one or more embodiments, field mapping is automated to reduce the time and / or amount of computing resources required to perform data migration between databases, data models, and / or systems. In one or more embodiments, field mapping is a hybrid solution that uses unsupervised machine learning and data insights (e.g., data knowledge) to intelligently learn mappings between databases, data models, and / or systems. In one or more embodiments, field mapping uses ground truth models, mapping models based on field names, mapping models based on field descriptions, and / or models for data features that are run sequentially to generate mapping results between databases, data models, and / or systems. In one or more embodiments, a mapping template for a first system (e.g., a target system), a data schema of a second system (e.g., a legacy system), and / or data from the first and second systems is used to recommend one or more top matching data fields between the first and second systems. In one or more embodiments, a mapping template for a first database, a data schema of the second database, and / or data from the first and second databases is used to recommend one or more top matching data fields between the first and second databases. In one or more embodiments, a mapping template for a first data model, a data schema of the second data model, and / or data from the first and second data models is used to recommend one or more top matching data fields between the first and second data models.

[0023] In one or more embodiments, a recurrent neural network is used to map the data to multidimensional word embeddings. In one or more embodiments, a network of gated recurrent units of the recurrent neural network is used to aggregate total spend. According to one or more embodiments, part commodity families are mapped to supplier commodity classifications based on part description data. Additionally or alternatively, in one or more embodiments, part commodity families are mapped to supplier commodity classifications based on purchase order description data. Additionally or alternatively, in one or more embodiments, part commodity families are mapped to supplier commodity classifications based on location data. Additionally or alternatively, in one or more embodiments, part commodity families are mapped to supplier commodity classifications based on expenditure type data. Additionally or alternatively, in one or more embodiments, part commodity families are mapped to supplier commodity classifications based on hierarchical data formatting techniques.

[0024] In one or more embodiments, a column name-based model and / or a column value-based model are used to facilitate mapping of data to multidimensional word embeddings. In embodiments, the column name-based model learns a vector representation of one or more defined target column names. The column name-based model also calculates similarity between source column names and one or more defined target column names. The one or more defined target column names are configured, for example, as full name strings or name abbreviations. In one or more embodiments, input to the column name-based model includes one or more source column names and / or one or more defined target column names. According to various embodiments, the one or more source column names are automatically identified from heterogeneous data sources. Feature generation for the column name-based model includes, for example, generating text embeddings for the column names of the source columns and / or target columns. Additionally, feature generation techniques for column name-based models include Term Frequency-Inverse Document Frequency (TF-IDF) + character-based n-grams, smooth inverse frequency (SIF), a library of learned word embeddings and / or text classification, universal sentence encoder, bidirectional encoder representations from transformer (BERT) embeddings, and / or one or more other feature generation techniques.

[0025] According to various embodiments, training the column name-based model involves the use of a hierarchical classification model including a Level 1 associated with predicting dataset categories and a Level 2 associated with predicting corresponding column names using the predicted dataset categories as features. According to various embodiments, training the column name-based model additionally or alternatively involves the use of a multi-class classification model associated with one or more decision tree algorithms configured to predict the most probable mappings to source columns. According to various embodiments, the column name-based model is trained on known target data. Furthermore, as more data becomes available, the additional data is used to, for example, include additional variations on data characteristics to enhance the performance of the column name-based model.

[0026] According to various embodiments, inference involving a column name-based model includes preparing data by generating features for column names in an input dataset. The trained version of the column name-based model is used to perform inference on new data obtained from heterogeneous data sources. For unmapped columns, one or more embodiments use cosine similarity to calculate a similarity score between pairs of source and target columns, e.g., using unsupervised learning.

[0027] The column value-based model provides a mapping technique based on column values ​​to generate a correct mapping. In embodiments, the column value-based model trains a text classifier using a Transformer model. In one or more embodiments, a pre-trained model, such as a RoBERT (base) model, is fine-tuned by using dense layers above the last layer of the neural network. In one or more embodiments, the neural network of the column value-based model is trained on a defined dataset having target column names and values. According to embodiments, the neural network of the column value-based model includes a set of Transformer encoder layers (e.g., 12 Transformer encoder layers), a set of hidden size representations (e.g., 768 hidden size representations), and / or a set of attention heads (e.g., 12 attention heads). The input to the column value-based model includes one or more column values ​​associated with the original source column names, source column values, and / or target column names. For example, in embodiments, the input to the column value-based model includes a list of column values ​​for all source columns. Furthermore, the output of the column value-based model includes a predicted target column mapping. In one or more embodiments, the raw text values ​​undergo tokenization, and / or the input is formatted (e.g., to obtain tokens, segments, positions, embeddings, padding, truncation, and / or attention masks) before being provided to the Transformer model. In one or more embodiments, a RoBERTa classification model is used with a single linear layer implemented on top of the model for classification associated with the text classifier. In one or more embodiments, once input data is provided to the column-value-based model, a pre-trained RoBERTa model and / or one or more additional untrained classification layers are trained based on the target dataset. In one or more embodiments, the neural network architecture for the column-value-based model includes providing the input column values ​​to character-level embeddings, providing data from the character-level embeddings to a Transformer, and providing data from the Transformer to a classifier.

[0028] In one or more embodiments, a scoring model is used to recommend an action based on different metrics from the recurrence history. In one or more embodiments, a user-interactive graphical user interface is generated. For example, in one or more embodiments, the graphical user interface renders a visual representation of the categorized purchase record data. In one or more embodiments, one or more notifications for the user device are generated based on the categorized purchase record data. In one or more embodiments, at least a portion of the recurrent neural network is retrained based on the categorized purchase record data.

[0029] Thus, by using one or more techniques disclosed herein, enterprise performance is optimized. For example, in one or more embodiments, one or more techniques disclosed herein are used to optimize spending (e.g., unclassified spending) related to one or more assets and / or services. In another example, in one or more embodiments, one or more techniques disclosed herein are used to optimize payment terms related to one or more assets and / or services. In another example, in one or more embodiments, one or more techniques disclosed herein are used to determine alternative suppliers of one or more assets and / or services. In another example, in one or more embodiments, one or more techniques disclosed herein are used to optimize shipping terms related to one or more assets and / or services. In another example, in one or more embodiments, one or more techniques disclosed herein are used to determine alternative target insights related to one or more assets and / or services. Furthermore, one or more techniques disclosed herein are used to improve field mapping for formatting disparate data associated with one or more data sources. Furthermore, one or more techniques disclosed herein are used to improve the quality of training data provided to an AI model. Furthermore, by using one or more techniques disclosed herein, improved insights into the unclassified data can be provided to users via improved visual indicators associated with a graphical user interface. For example, by using one or more techniques disclosed herein, additional and / or improved insights can be achieved across an entire data set compared to the capabilities of conventional techniques. Additionally, the performance of processing systems associated with data analysis can be improved by using one or more techniques disclosed herein. For example, the number of computing resources, storage requirements, and / or the number of errors associated with data analysis can be reduced by using one or more techniques disclosed herein.

[0030] Figure 1 illustrates an exemplary networked computing system environment 100 according to the present disclosure. As shown in Figure 1, the networked computing system environment 100 is comprised of multiple tiers, including a cloud tier 105, a network tier 110, and an edge tier 115. Components of the edge 115 communicate with components of the cloud 105 via the network 110, as described in further detail below.

[0031] In various embodiments, network 110 is any suitable network or combination of networks supporting any appropriate protocols suitable for communicating data to and from components of cloud 105 and various other components (e.g., components of edge 115) in networked computing system environment 100. According to various embodiments, network 110 includes a public network (e.g., the Internet), a private network (e.g., an internal organizational network), or a combination of public and / or private networks. According to various embodiments, network 110 is configured to provide communication between the various components shown in FIG. 1 . According to various embodiments, network 110 includes one or more networks that connect devices and / or components in a network layout to enable communication between the devices and / or components. For example, in one or more embodiments, network 110 is implemented as the Internet, a wireless network, a wired network (e.g., Ethernet), a local area network (LAN), a wide area network (WAN), Bluetooth, near field communication (NFC), or any other type of network that provides communication between one or more components of a network layout. In some embodiments, network 110 is implemented using a cellular network, satellite, licensed radio, or a combination of cellular, satellite, licensed radio, and / or unlicensed radio networks.

[0032] Components of cloud 105 include one or more computer systems 120 that form a so-called “Internet of Things” or “IoT” platform 125. It should be appreciated that “IoT platform” is an optional term describing a platform that connects any type of Internet-connected device and should not be construed as limiting the types of computing systems that can be used within IoT platform 125. In particular, in various embodiments, computer system 120 includes any type or quantity of one or more processors and one or more data storage devices that include memory for storing and executing applications or software modules of networked computing system environment 100. In one embodiment, the processor and data storage device are embodied in server-class hardware, such as an enterprise-level server. For example, in embodiments, the processor and data storage device comprise any type of application server, communication server, web server, supercomputing server, database server, file server, mail server, proxy server, and / or virtual server, or a combination thereof. Further, one or more processors are configured to access the memory and execute processor-readable instructions that, when executed by the processor, configure the processor to perform a plurality of functions of the networked computing system environment 100.

[0033] Computer system 120 further includes one or more software components of IoT platform 125. For example, in one or more embodiments, the software components of computer system 120 include one or more software modules for communicating with user devices and / or other computing devices over network 110. For example, in one or more embodiments, the software components include one or more modules 141, models 142, engines 143, databases 144, services 145, and / or applications 146, which may be stored within / by computer system 120 (e.g., stored in memory), as described in more detail with respect to FIG. 2 below. According to various embodiments, one or more processors are configured to utilize one or more of modules 141, models 142, engines 143, databases 144, services 145, and / or applications 146 when performing various methods described in this disclosure.

[0034] Thus, in one or more embodiments, computer system 120 may execute a cloud computing platform (e.g., IoT platform 125) with scalable resources for computation and / or data storage, and may execute one or more applications on the cloud computing platform to perform various computer-implemented methods described in this disclosure. In some embodiments, some of modules 141, models 142, engines 143, databases 144, services 145, and / or applications 146 are combined to form fewer modules, models, engines, databases, services, and / or applications. In some embodiments, some of modules 141, models 142, engines 143, databases 144, services 145, and / or applications 146 are separated into separate, larger modules, models, engines, databases, services, and / or applications. In some embodiments, some of modules 141, models 142, engines 143, databases 144, services 145, and / or applications 146 are removed and others are added.

[0035] Computer system 120 is configured to receive data from other components of networked computing system environment 100 (e.g., components of edge 115) via network 110. Computer system 120 is further configured to generate results utilizing the received data. According to various embodiments, information indicative of the results is transmitted over network 110 to a user via a user computing device. In some embodiments, computer system 120 is a server system that provides one or more services, including providing the received data and / or information indicative of the results to a user. According to various embodiments, computer system 120 is part of an entity, including any type of company, organization, or institution, that performs one or more IoT services. In some examples, the entity is an IoT platform provider.

[0036] Components of the edge 115 include one or more enterprises 160a-160n, each including one or more edge devices 161a-161n and one or more edge gateways 162a-162n. For example, a first enterprise 160a includes a first edge device 161a and a first edge gateway 162a, a second enterprise 160b includes a second edge device 161b and a second edge gateway 162b, and an nth enterprise 160n includes an nth edge device 161n and an nth edge gateway 162n. As used herein, enterprises 160a-160n represent any type of entity, facility, or vehicle, such as, for example, a company, a department, a building, a manufacturing plant, a warehouse, a real estate facility, a research lab, an aircraft, a spacecraft, an automobile, a ship, a boat, a military vehicle, an oil and gas facility, or any other type of entity, facility, and / or vehicle that includes any number of local devices.

[0037] According to various embodiments, edge devices 161a-161n represent any of a variety of different types of devices that may be used within enterprises 160a-160n. Edge devices 161a-161n are any type of device configured to access network 110 or accessed by other devices through network 110, such as via edge gateways 162a-162n. According to various embodiments, edge devices 161a-161n are "IoT devices," which include any type of network-connected (e.g., internet-connected) device. For example, in one or more embodiments, edge devices 161a-161n include sensors, actuators, processors, computers, valves, pumps, ducts, vehicle components, cameras, displays, doors, windows, security components, HVAC components, factory facilities, and / or any other devices connected to network 110 to collect, transmit, and / or receive information. Each edge device 161a-161n includes or otherwise communicates with one or more controllers for selectively controlling the respective edge device 161a-161n and / or for sending / receiving information between the edge device 161a-161n and the cloud 105 via the network 110. Referring to FIG. 2 , in one or more embodiments, the edge 115 includes operational technology (OT) systems 163a-163n and information technology (IT) applications 164a-164n for each enterprise 161a-161n. The OT systems 163a-163n include hardware and software for detecting and / or causing changes through direct monitoring and / or control of industrial equipment (e.g., edge devices 161a-161n), assets, processes, and / or events. The IT applications 164a-164n include network, storage, and computing resources for generating, managing, storing, and distributing data throughout and between organizations.

[0038] The edge gateways 162a-162n include devices for facilitating communication between the edge devices 161a-161n and the cloud 105 over the network 110. For example, the edge gateways 162a-162n include one or more communication interfaces for communicating with the edge devices 161a-161n and with the cloud 105 over the network 110. According to various embodiments, the communication interfaces of the edge gateways 162a-162n include one or more cellular, Bluetooth, WiFi, near field communication, Ethernet, or other suitable communication devices for transmitting and receiving information. According to various embodiments, multiple communication interfaces are included in each gateway 162a-162n to provide multiple forms of communication between the edge devices 161a-161n, the gateways 162a-162n, and the cloud 105 over the network 110. For example, in one or more embodiments, communication with edge devices 161a-161n and / or network 110 is achieved through wireless communication (e.g., WiFi, wireless communication, etc.) and / or wired data connections (e.g., Universal Serial Bus, on-board diagnostic systems, etc.) or other communication modes, such as a local area network (LAN), a wide area network (WAN) such as the Internet, a telecommunications network, a data network, or any other type of network.

[0039] According to various embodiments, the edge gateways 162a-162n also include processors and memory for storing and executing program instructions to facilitate data processing. For example, in one or more embodiments, the edge gateways 162a-162n are configured to receive data from the edge devices 161a-161n and process the data before transmitting the data to the cloud 105. Thus, in one or more embodiments, the edge gateways 162a-162n include one or more software modules or components for providing data processing services and / or other services or methods of the present disclosure. With reference to FIG. 2, each edge gateway 162a-162n includes an edge service 165a-165n and an edge connector 166a-166n. According to various embodiments, the edge service 165a-165n includes hardware and software components for processing data from the edge devices 161a-161n. According to various embodiments, the edge connectors 166a-166n include hardware and software components for facilitating communication between the edge gateways 162a-162n and the cloud 105 via the network 110, as detailed above. In some cases, any of the edge devices 161a-n, edge connectors 166a-n, and edge gateways 162a-n may have their functions combined, omitted, or separated into any combination of devices. In other words, the edge devices and their connectors and gateways are not necessarily separate devices.

[0040] 2 shows a schematic block diagram of a framework 200 of an IoT platform 125 according to the present disclosure. The IoT platform 125 of the present disclosure is a platform for enterprise performance management that uses real-time, accurate models and visual analytics to deliver rational, actionable recommendations for sustained peak performance of an enterprise 160a-160n. The IoT platform 125 is a portable, extensible platform that can be deployed in any cloud or data center environment to provide a top-down view of the entire enterprise, displaying the status of processes, assets, people, and safety. Additionally, the IoT platform 125 supports end-to-end capabilities, using the framework 200, described in further detail below, to run digital twins on process data and convert the output into actionable insights.

[0041] 2 , the framework 200 of the IoT platform 125 includes several layers, including, for example, an IoT layer 205, an enterprise integration layer 210, a data pipeline layer 215, a data insights layer 220, an application services layer 225, and an application layer 230. The IoT platform 125 also includes a core services layer 235 and an extensible object model (EOM) 250 that includes one or more knowledge graphs 251. The layers 205-235 further include various software components that together form each layer 205-235. For example, in one or more embodiments, each layer 205-235 includes one or more of: a module 141, a model 142, an engine 143, a database 144, a service 145, an application 146, or a combination thereof. In some embodiments, the layers 205-235 are combined to form fewer layers. In some embodiments, some of the layers 205-235 are separated into separate, larger layers. In some embodiments, some of the layers 205-235 may be removed and other layers may be added.

[0042] The IoT platform 125 is a model-driven architecture. Thus, an extensible object model 250 communicates with each layer 205-230 and contextualizes the site data of the enterprises 160a-160n using an extensible object model (or "asset model") and a knowledge graph 251 in which the equipment (e.g., edge devices 161a-161n) and processes of the enterprises 160a-160n are modeled. The knowledge graph 251 of the EOM 250 is configured to store the model in a central location. The knowledge graph 251 defines a collection of nodes and links that describe the real-world connections that enable smart systems. As used herein, a knowledge graph 251 (i) describes real-world entities (e.g., edge devices 161a-161n) and their interrelationships organized in a graphical interface, (ii) defines possible classes and relationships of entities in a schema, (iii) allows arbitrary entities to be related to each other, and (iv) targets various topic domains. In other words, knowledge graph 251 defines a large network of entities (e.g., edge devices 161a-161n), the semantic types of the entities, the properties of the entities, and the relationships between the entities. Thus, knowledge graph 251 describes a network of “things” associated with a particular domain or enterprise or organization. Knowledge graph 251 is not limited to abstract concepts and relationships but can also include instances of objects, such as documents and datasets. In some embodiments, knowledge graph 251 includes a resource description framework (RDF) graph. As used herein, an “RDF graph” is a graph data model that formally describes the semantics or meaning of information. RDF graphs also represent metadata (e.g., data that describes data). According to various embodiments, knowledge graph 251 also includes a semantic object model. A semantic object model is a subset of knowledge graph 251 that defines the semantics of knowledge graph 251.For example, the semantic object model defines the schema of the knowledge graph 251 .

[0043] As used herein, EOM 250 is a collection of application programming interfaces (APIs) that allow a seeded semantic object model to be extended. For example, EOM 250 of the present disclosure allows a customer's knowledge graph 251 to be constructed according to constraints expressed in the customer's semantic object model. Thus, knowledge graph 251 is generated by a customer (e.g., a business or organization) to create a model of edge devices 161 a-161 n of the enterprise 160 a-160 n, and knowledge graph 251 is input into EOM 250 to visualize the model (e.g., nodes and links).

[0044] The model describes the assets (e.g., nodes) of the enterprise (e.g., edge devices 161a-161n) and describes the relationships between the assets and other components (e.g., links). The model also describes the schema (e.g., describes what the data is), and therefore the model is self-validating. For example, in one or more embodiments, the model describes the types of sensors attached to any given asset (e.g., edge devices 161a-161n) and the type of data being sensed by each sensor. According to various embodiments, a key performance indicator (KPI) framework is used to bind the properties of the assets in the extensible object model 250 to the inputs of the KPI framework. Thus, the IoT platform 125 is an extensible, model-driven, end-to-end stack that includes bidirectional model synchronization and secure data exchange between the edge 115 and the cloud 105, metadata-driven data processing (e.g., rules, calculations, and aggregations), and model-driven visualizations and applications. As used herein, "extensible" refers to the ability to extend the data model to include new properties / columns / fields, new classes / tables, and new relationships. Thus, the IoT platform 125 is extensible with respect to edge devices 161a-161n and the applications 146 that process those devices 161a-161n. For example, when new edge devices 161a-161n are added to the enterprise 160a-160n system, the new devices 161a-161n automatically appear in the IoT platform 125 so that the corresponding applications 146 understand and use data from the new devices 161a-161n.

[0045] In some cases, asset templates are used to facilitate the configuration of edge device 161a-n instances within a model using a common structure. Asset templates define typical properties for edge devices 161a-n at a given enterprise 160a-n for a particular type of device. For example, an asset template for a pump may include modeling a pump with inlet and outlet pressures, speeds, flow rates, etc. Templates may also include hierarchies or derived types of edge devices 161a-n to accommodate variations on the basic type of device 161a-n. For example, a reciprocating pump is a specialization of the basic pump type and includes additional properties in the template. Instances of edge devices 161a-n within a model are configured to match the actual physical devices of the enterprise 160a-n using templates to define the expected attributes of the device 161a-n. Each attribute is configured as a static value (e.g., capacity is 1000 BPH) or with a reference to a time-series tag that provides the value. The knowledge graph 250 can automatically map tags to attributes based on naming conventions, parsing, and matching of tag and attribute descriptions and / or by comparing the behavior of the time series data with expected behavior.

[0046] The modeling phase includes an onboarding process for synchronizing models between the edge 115 and the cloud 105. For example, in one or more embodiments, the onboarding process includes a simple onboarding process, a complex onboarding process, and / or a standardized rollout process. The simple onboarding process includes the knowledge graph 250 receiving raw model data from the edge 115 and running a context discovery algorithm to generate a model. The context discovery algorithm reads the context of the edge naming conventions of the edge devices 161a-161n and determines what the naming conventions refer to. For example, in one or more embodiments, the knowledge graph 250 receives "TMP" during the modeling phase and determines that "TMP" relates to "temperature." The generated model is then published. The complex onboarding process includes the knowledge graph 250 receiving raw model data, receiving location history data, and receiving fieldwork data. According to various embodiments, the knowledge graph 250 then runs a context discovery algorithm using these inputs. According to various embodiments, the generated model is edited, and then the model is published. The standardized rollout process involves manually defining a standard model in the cloud 105 and pushing the model to the edge 115.

[0047] The IoT layer 205 includes one or more components for device management, data ingestion, and / or command / control of the edge devices 161a-161n. The components of the IoT layer 205 allow data to be ingested into or otherwise received by the IoT platform 125 from various sources. For example, in one or more embodiments, data is ingested from the edge devices 161a-161n via a process historian or laboratory information management system. The IoT layer 205 communicates with edge connectors 165a-165n located on the edge gateways 162a-162n over the network 110, which securely transmit data to the IoT platform 205. In some embodiments, only authorized data is sent to the IoT platform 125, and the IoT platform 125 only accepts data from authorized edge gateways 162a-162n and / or edge devices 161a-161n. According to various embodiments, data is transmitted from the edge gateways 162a-162n to the IoT platform 125 via direct streaming and / or batch delivery. Additionally, after any network or system outage, data transfer resumes once communication is re-established, and any data lost during the outage is backfilled from the source system or from the cache of the IoT platform 125. According to various embodiments, the IoT layer 205 also includes components for accessing time series, alarm and event, and transaction data via various protocols.

[0048] The enterprise integration layer 210 includes one or more components for events / messaging, file upload, and / or REST / OData. The components of the enterprise integration layer 210 enable the IoT platform 125 to communicate with third-party cloud applications 211, such as any applications operated by an enterprise in connection with its edge devices. For example, the enterprise integration layer 210 connects with enterprise databases such as a guest database, a customer database, a financial database, a patient database, etc. The enterprise integration layer 210 provides third parties with standard application programming interfaces (APIs) for accessing the IoT platform 125. The enterprise integration layer 210 also enables the IoT platform 125 to communicate with the OT systems 163a-163n and IT applications 164a-164n of the enterprises 160a-160n. Thus, the enterprise integration layer 210 enables the IoT platform 125 to receive data from third-party applications 211 instead of or in combination with receiving data directly from the edge devices 161a-161n.

[0049] The data pipeline layer 215 includes one or more components for data cleansing / enrichment, data transformation, data calculation / aggregation, and / or APIs for data streams. Thus, in one or more embodiments, the data pipeline layer 215 pre-processes and / or performs initial analysis on received data. The data pipeline layer 215 performs advanced data cleansing routines, including, for example, data correction, mass balancing, data adjustment, component balancing, and simulation, to ensure the desired information is used as the basis for further processing. The data pipeline layer 215 also performs advanced, high-speed calculations. For example, the cleansed data is run through an enterprise-specific digital twin. According to various embodiments, the enterprise-specific digital twin includes a reliability advisor that includes a process model to determine current operation and a fault model to trigger any early detection and determine appropriate resolution. According to various embodiments, the digital twin also includes an optimization advisor that integrates real-time economic data with real-time process data to select the correct feed for the process and determine optimal process conditions and product yields.

[0050] According to various embodiments, the data pipeline layer 215 uses models and templates to define calculations and analytics. Additionally or alternatively, according to various embodiments, the data pipeline layer 215 uses models and templates to define how calculations and analytics relate to assets (e.g., edge devices 161a-161n). For example, in embodiments, a pump template defines a pump efficiency calculation so that a standard efficiency calculation is automatically performed for a pump each time the pump is configured. A calculation model defines various types of calculations, the type of engine on which the calculation should be performed, input and output parameters, pre-processing requirements and prerequisites, schedules, etc. According to various embodiments, the actual calculation or analysis logic may be defined or referenced within the template. Thus, according to various embodiments, a calculation model is used to describe and control the execution of various different process models. According to various embodiments, a calculation template is linked to an asset template so that when an asset (e.g., edge device 161a-161n) instance is created, any associated calculation instances are also created, with input and output parameters linked to the appropriate attributes of the asset (e.g., edge device 161a-161n).

[0051] According to various embodiments, the IoT platform 125 supports a variety of different analytical models, including, for example, first principles models, empirical models, engineering models, user-defined models, machine learning models, built-in functions, and / or any other type of analytical model. While failure models and predictive maintenance models are described herein as examples, any type of model may be applied.

[0052] Fault models are used to compare the current and predicted performance of the enterprises 160a-160n to identify problems or opportunities and their potential causes or drivers. The IoT platform 125 includes a rich set of hierarchical symptom-fault models for identifying abnormal conditions and their potential consequences. For example, in one or more embodiments, the IoT platform 125 drills down from high-level conditions to understand contributing factors and determine the potential impact that lower-level conditions may have. There may be multiple fault models for a given enterprise 160a-160n that look at different aspects of processes, equipment, controls, and / or operations. According to various embodiments, each fault model identifies problems and opportunities in their domain and may even look at the same core problem from different perspectives. According to various embodiments, an overall fault model is layered on top to synthesize the different perspectives from each fault model into an overall assessment of the situation and pinpoint the true root cause.

[0053] According to various embodiments, once a fault or opportunity is identified, the IoT platform 125 provides recommendations regarding the optimal corrective action to take. Initially, recommendations are based on specialized knowledge pre-programmed into the system by process and equipment experts. The recommendation service module presents this information in a consistent manner regardless of source and supports workflows for tracking, completing, and documenting recommendation follow-up. According to various embodiments, recommendation follow-up is used to improve the overall knowledge of the system over time as existing recommendations are substantiated (or not) or as new causal relationships are learned by the user and / or analytics.

[0054] According to various embodiments, models are used to accurately predict what will happen before it happens and to interpret the state of the installed base. Thus, IoT platform 125 enables operators to quickly initiate maintenance actions when an anomaly occurs. According to various embodiments, the digital twin architecture of IoT platform 125 uses various modeling techniques. According to various embodiments, modeling techniques include, for example, rigorous models, fault detection and diagnostics (FDD), descriptive models, predictive maintenance, prescriptive maintenance, process optimization, and / or any other modeling technique.

[0055] According to various embodiments, rigorous models are converted from process design simulations. In this way, process designs are integrated with feed conditions and production requirements. Process changes and technology improvements provide business opportunities that enable more effective maintenance schedules and resource deployment in the context of production needs. Fault detection and diagnosis include generalized rule sets that are specified based on industry experience and domain knowledge and can be easily integrated and used to work with equipment models. According to various embodiments, descriptive models identify problems, and predictive models determine possible damage levels and maintenance options. According to various embodiments, descriptive models include models for defining operating windows for edge devices 161a-161n.

[0056] Predictive maintenance involves predictive analytics models developed based on rigorous and statistical models, such as principal component analysis (PCA) and partial least squares (PLS). According to various embodiments, machine learning methods are applied to training models for failure prediction. According to various embodiments, predictive maintenance utilizes FDD-based algorithms to continuously monitor individual control and equipment performance. Predictive modeling is then applied to selected condition indicators that degrade over time. Prescriptive maintenance involves determining optimal maintenance options and when to perform them based on actual conditions rather than time-based maintenance schedules. According to various embodiments, prescriptive analytics selects the correct solution based on a company's capital, operational, and / or other requirements. Process optimization involves determining optimal conditions by adjusting set points and schedules. Optimized set points and schedules can be directly communicated to the underlying controller, allowing for the automatic closing of the loop from analysis to control.

[0057] The data insights layer 220 includes one or more components for a time-series database (TDSB), a relational / document database, a data lake, blobs, files, images, and videos, and / or APIs for data querying. According to various embodiments, when raw data is received at the IoT platform 125, it is stored in warm storage (e.g., a TSDB) as time-series tags or events to support interactive querying, and in cold storage for archival purposes. According to various embodiments, the data is sent to a data lake for offline analytical deployment. According to various embodiments, the data pipeline layer 215 accesses the data stored in the databases of the data insights layer 220 and performs analysis, as described in detail above.

[0058] The application services layer 225 includes one or more components for a rules engine, workflow / notifications, a KPI framework, insights (e.g., actionable insights), decisions, recommendations, machine learning, and / or APIs for application services. The application services layer 225 enables the construction of applications 146a-d. The application layer 230 includes one or more applications 146a-d of the IoT platform 125. For example, according to various embodiments, the applications 146a-d include a building application 146a, a plant application 146b, an aviation application 146c, and other enterprise applications 146d. According to various embodiments, the applications 146a-d include a general-purpose application 146 for portfolio management, asset management, autonomous control, and / or any other custom application. According to various embodiments, portfolio management includes a KPI framework and a flexible user interface (UI) builder. According to various embodiments, asset management includes asset performance and asset health. According to various embodiments, autonomous control includes energy optimization and / or predictive maintenance. As detailed above, according to various embodiments, the general-purpose application 146 is extensible such that each application 146 is configurable for different types of enterprises 160a-160n (e.g., building application 146a, plant application 146b, aero application 146c, and other enterprise application 146d).

[0059] The application layer 230 also enables visualization of the performance of the enterprises 160a-160n. For example, dashboards provide high-level overviews with drill-downs to support deeper investigation. Recommendation summaries empower users with prioritized actions to address current or potential problems and opportunities. Data analysis tools support ad-hoc data exploration to facilitate troubleshooting and process improvement.

[0060] The core services layer 235 includes one or more services of the IoT platform 125. According to various embodiments, the core services 235 include data visualization, data analysis tools, security, scaling, and monitoring. According to various embodiments, the core services 235 also include services for tenant provisioning, single login / common portal, self-service management, UI libraries / UI tiles, identity / access / entitlements, logging / monitoring, usage metering, API gateway / dev portal, and IoT platform 125 streams.

[0061] 3 illustrates a system 300 providing an exemplary environment in accordance with one or more described features of one or more embodiments of the present disclosure. According to an embodiment, the system 300 includes a data optimization computer system 302 for facilitating the practical application of data analytics and / or digital transformation techniques to perform optimizations related to enterprise performance management. In one or more embodiments, the data optimization computer system 302 facilitates the practical application of machine learning techniques to perform optimizations related to enterprise performance management. In one or more embodiments, the data optimization computer system 302 analyzes ingested, cleaned, and / or aggregated data from one or more information technology data sources to provide cost savings and / or efficiency insights for enterprise systems.

[0062] In embodiments, the data-optimized computer system 302 is a server system (e.g., a server device) that facilitates a data analysis platform between one or more computing devices and one or more data sources. In one or more embodiments, the data-optimized computer system 302 is a device having one or more processors and memory. In one or more embodiments, the data-optimized computer system 302 is a computer system from computer system 120. For example, in one or more embodiments, the data-optimized computer system 302 is implemented via cloud 105. The data optimization computer system 302 may also relate to one or more technologies, such as, for example, enterprise technology, data analytics technology, digital transformation technology, cloud computing technology, cloud database technology, server technology, network technology, wireless communication technology, natural language processing technology, machine learning technology, artificial intelligence technology, digital processing technology, electronic device technology, computer technology, industrial technology, Industrial Internet of Things (IoT) technology, supply chain analytics technology, aircraft technology, construction technology, cybersecurity technology, navigation technology, asset visualization technology, oil and gas technology, petrochemical technology, refinery technology, process plant technology, procurement technology, and / or one or more other technologies.

[0063] Additionally, the data optimization computer system 302 improves one or more technologies, such as enterprise technology, data analytics technology, digital transformation technology, cloud computing technology, cloud database technology, server technology, network technology, wireless communication technology, natural language processing technology, machine learning technology, artificial intelligence technology, digital processing technology, electronic device technology, computer technology, industrial technology, Industrial Internet of Things (IoT) technology, supply chain analysis technology, aircraft technology, construction technology, cybersecurity technology, navigation technology, asset visualization technology, oil and gas technology, petrochemical technology, refinery technology, process plant technology, procurement technology, and / or one or more other technologies. In one implementation, the data optimization computer system 302 improves the performance of a computing device. For example, in one or more embodiments, the data optimization computer system 302 improves the processing efficiency of a computing device (e.g., a server), reduces the power consumption of a computing device (e.g., a server), improves the quality of data provided by a computing device (e.g., a server), etc.

[0064] Data optimization computer system 302 includes a data mapping component 304, an artificial intelligence component 306, and / or an action component 308. Furthermore, in certain embodiments, data optimization computer system 302 includes a processor 310 and / or memory 312. In certain embodiments, one or more aspects of data optimization computer system 302 (and / or other systems, devices, and / or processes disclosed herein) constitute executable instructions embodied in a computer-readable storage medium (e.g., memory 312). For example, in embodiments, memory 312 stores computer-executable components and / or executable instructions (e.g., program instructions). Furthermore, processor 310 facilitates the execution of the computer-executable components and / or executable instructions (e.g., program instructions). In an exemplary embodiment, processor 310 is configured to execute instructions stored in memory 312 or otherwise accessible to processor 310.

[0065] The processor 310 is a hardware entity (e.g., physically embodied in a circuit) capable of performing operations according to one or more embodiments of the present disclosure. Alternatively, in embodiments in which the processor 310 is embodied as an executor of software instructions, the software instructions configure the processor 310 to perform one or more algorithms and / or operations described herein in response to the software instructions being executed. In embodiments, the processor 310 is a single-core processor, a multi-core processor, multiple processors internal to the data optimization computer system 302, a remote processor (e.g., a processor implemented on a server), and / or a virtual machine. In particular embodiments, the processor 310 communicates with the memory 312, the data mapping component 304, the artificial intelligence component 306, and / or the action component 308 via a bus to facilitate, for example, the transmission of data between the processor 310, the memory 312, the data mapping component 304, the artificial intelligence component 306, and / or the action component 308. The processor 310 may be embodied in several different ways and, in particular embodiments, may include one or more processing units configured to operate independently. Additionally or alternatively, in one or more embodiments, processor 310 includes one or more processors configured in tandem via a bus to enable independent execution of instructions, pipelining of data, and / or multi-threaded execution of instructions.

[0066] The memory 312 is non-transitory, e.g., includes one or more volatile memories and / or one or more non-volatile memories. In other words, in one or more embodiments, the memory 312 is an electronic storage device (e.g., a computer-readable storage medium). The memory 312 is configured to store information, data, content, one or more applications, one or more instructions, etc., to enable the data-optimized computer system 302 to perform various functions in accordance with one or more embodiments disclosed herein. As used herein in this disclosure, terms such as "component," "system," etc., are computer-related entities. For example, a "component," "system," etc., disclosed herein may be either hardware, software, or a combination of hardware and software. By way of example, a component may be, but is not limited to, a process running on a processor, a processor, a circuit, an executable component, a thread of instructions, a program, and / or a computer entity.

[0067] In embodiments, the data optimized computer system 302 (e.g., the data mapping component 304 of the data optimized computer system 302) receives the heterogeneous data 314. In one or more embodiments, the data optimized computer system 302 (e.g., the data mapping component 304 of the data optimized computer system 302) receives the heterogeneous data 314 from one or more data sources 316. In particular embodiments, at least one data source from the one or more data sources 316 incorporates encryption functionality that facilitates encryption of one or more portions of the heterogeneous data 314. In particular embodiments, the one or more data sources 316 are one or more IT data sources. Additionally, in one or more embodiments, the data optimized computer system 302 (e.g., the data mapping component 304 of the data optimized computer system 302) receives the heterogeneous data 314 over the network 110. In one or more embodiments, network 110 is a Wi-Fi network, a near field communication (NFC) network, a Worldwide Interoperability for Microwave Access (WiMAX) network, a personal area network (PAN), a short-range wireless network (e.g., a Bluetooth network), an infrared wireless (e.g., IrDA) network, an ultra-wideband (UWB) network, an inductive radio transmission network, and / or another type of network. In one or more embodiments, one or more data sources 316 are associated with components of edge 115, such as, for example, one or more enterprises 160a-160n. In one or more embodiments, one or more data sources 316 are similar but non-uniform data sources. For example, in embodiments, one or more data sources 316 are procurement data sources in different subparts of an enterprise system (e.g., a procurement system and a financial system, a sales system and a procurement system, etc.).

[0068] Disparate data 314 may include, for example, uncategorized data, uncategorized data elements, uncategorized data entities, and / or other uncategorized information. In certain embodiments, disparate data 314 may additionally include categorized data (e.g., previously categorized data). Furthermore, in one or more embodiments, disparate data 314 may include one or more data fields (e.g., one or more fillable fields). In one or more embodiments, a data field associated with disparate data 314 may include a data element, be formatted with a data element, and / or be tagged with a data element. Alternatively, in one or more embodiments, a data field associated with disparate data 314 may be an incomplete data field formatted without a data element. In one or more embodiments, the disparate data 314 includes transaction data (e.g., uncategorized transaction data), purchase record data (e.g., uncategorized purchase record data), invoice data (e.g., uncategorized invoice data), purchase order data (e.g., uncategorized purchase order data), supplier data (e.g., uncategorized supplier data), contract data (e.g., uncategorized contract data), process data (e.g., uncategorized process data), industrial data (e.g., uncategorized industrial data), asset data (e.g., uncategorized asset data), shipping data (e.g., uncategorized shipping data), sensor data (e.g., uncategorized sensor data), location data (e.g., uncategorized location data), user data (e.g., uncategorized purchase record data), and / or other data (e.g., other uncategorized data). In one example, at least a portion of the disparate data 314 includes data associated with one or more dynamically modifiable electronic purchasing agreements. In another example, at least a portion of the invoice data associated with disparate data 314 includes a purchase order number, an invoice number, a supplier identifier, a payment term, an invoice amount, a supplier hierarchy level identifier, and / or other invoice information. In another example, at least a portion of the purchasing data associated with disparate data 314 includes a purchase order number, a supplier identifier, a purchase order line item, a purchase order balance, a purchase order terms, a part number, a product item family, a part description, and / or other purchase order information.

[0069] In one or more embodiments, the data mapping component 304 aggregates the disparate data 314 from one or more data sources 316. For example, in one or more embodiments, the data mapping component 304 can aggregate the disparate data 314 into a data lake 318. In one or more embodiments, the data lake 318 is a centralized repository (e.g., a single data lake) that stores the unstructured and / or structured data included in the disparate data 314. In one or more embodiments, the data mapping component 304 repeatedly updates the data in the data lake 318 at one or more predetermined intervals. For example, in one or more embodiments, the data mapping component 304 stores new and / or modified data associated with one or more data sources 316. In one or more embodiments, the data mapping component 304 repeatedly scans the one or more data sources 316 to determine new data for storage in the data lake 318.

[0070] In one or more embodiments, the data mapping component 304 formats one or more portions of the disparate data 314. For example, in one or more embodiments, the data mapping component 304 provides a formatted version of the disparate data 314. In embodiments, the formatted version of the disparate data 314 is formatted using one or more defined formats. The defined format is, for example, a structure of data fields. In one embodiment, the defined format is predetermined. For example, in one or more embodiments, a predominant type structure (e.g., predominant type format, predominant type procurement format, etc.) may be used as a template for future use. In another embodiment, the defined format is determined based on an analysis of the disparate data 314 (e.g., in response to a majority of the disparate data 314 being received). In various embodiments, the formatted version of the disparate data 314 is stored in the data lake 318.

[0071] In one or more embodiments, the data mapping component 304 identifies one or more distinct data fields in the disparate data 314 that describe the corresponding subject matter. For example, in one or more embodiments, the data mapping component 304 identifies one or more distinct data fields in the disparate data 314 that describe the corresponding vendor name. In another example, the mapping component 304 identifies one or more distinct data fields in the disparate data 314 that describe the corresponding payment terms. In one or more embodiments, the data mapping component 304 determines one or more incomplete data fields in the disparate data 314 that correspond to the identified one or more distinct data fields. In accordance with a determination that the one or more determined incomplete data fields correspond to the identified one or more distinct data fields, in one or more embodiments, the data mapping component 304 adds data from the identified data fields to the incomplete data fields in the disparate data 314. In one or more embodiments, the data mapping component 304 assigns one or more tags and / or metadata to the disparate data 314. In one or more embodiments, the data mapping component 304 extracts data from the disparate data 314 using one or more natural language processing techniques. In one or more embodiments, the data mapping component 304 determines one or more data elements, one or more words, and / or one or more phrases associated with the disparate data 314. In one or more embodiments, the data mapping component 304 predicts data for a data field based on particular intent associated with different data elements, words, and / or phrases associated with the disparate data 314. For example, in embodiments, the data mapping component 304 predicts data for a first data field associated with the transaction data based on particular intent associated with different data elements, words, and / or phrases associated with other transaction data stored in the disparate data 314.In another example related to another embodiment, the data mapping component 304 predicts data for a first data field associated with the industrial data based on specific intent associated with different data elements, words, and / or phrases associated with other industrial data stored in the disparate data 314. In one or more embodiments, the data mapping component 304 identifies and / or groups data types associated with the disparate data 314 based on a hierarchical data format. In one or more embodiments, the data mapping component 304 facilitates data mapping associated with the disparate data 314 using batch processing, data string concatenation, data type identification, data merging, data reading, and / or data writing. In one or more embodiments, the data mapping component 304 performs feature processing to remove one or more defined characters (e.g., special characters), feature processing to tokenize one or more strings, feature processing to remove one or more defined words (e.g., one or more stop words), feature processing to remove one or more single-character tokens, and / or other feature processing on the disparate data 314. In one or more embodiments, the data mapping component 304 groups data from the disparate data 314 based on corresponding features of the data. In one or more embodiments, the data mapping component 304 groups data from the disparate data 314 based on corresponding identifiers of the data (e.g., matching part commodity families). In one or more embodiments, the data mapping component 304 groups data from the disparate data 314 based on similarity scores and / or calculated distances between different data in the disparate data 314 using one or more locality-sensitive hashing techniques.

[0072] In one or more embodiments, the data mapping component 304 organizes the formatted version of the disparate data 314 based on an ontology tree structure. For example, in one or more embodiments, the data mapping component 304 organizes the formatted version of the disparate data 314 in an ontology tree structure using hierarchical data formatting techniques. In embodiments, the ontology tree structure captures relationships between different data in the disparate data 314 based on a hierarchy of nodes and connections between different data in the disparate data 314. In embodiments, nodes in the ontology tree structure correspond to data elements, and connections in the ontology tree structure represent relationships between nodes (e.g., data elements) in the ontology tree structure. In one or more embodiments, the data mapping component 304 traverses the ontology tree structure to traverse associations between aspects of the disparate data 314. In one or more embodiments, the data mapping component 304 compares different data sources in the one or more data sources 316 and / or data from different data sources in the one or more data sources 316 based on the ontology tree structure.

[0073] In one or more embodiments, the data mapping component 304 generates one or more features associated with a format structure of the heterogeneous data 314. For example, in one or more embodiments, the data mapping component 304 generates one or more features associated with one or more defined formats for the format structure. The format structure may be, for example, a target format structure for the heterogeneous data 314. In one or more embodiments, the format structure may be a format structure for one or more portions of the data lake 318. In embodiments, the one or more features include one or more data field features for the format structure. For example, in embodiments, the one or more features include one or more column name features for the format structure. Additionally or alternatively, in embodiments, the one or more features include one or more column value features for the format structure. However, it should be appreciated that the one or more features may additionally or alternatively include one or more other types of features associated with the format structure. In particular embodiments, the one or more features generated by the data mapping component 304 include one or more text embeddings for column names associated with the format structure. For example, in certain embodiments, the one or more features generated by the data mapping component 304 include one or more text embeddings for column names associated with source column names and / or target column names for one or more portions of the heterogeneous data 314. Additionally or alternatively, in certain embodiments, the one or more features generated by the data mapping component 304 include one or more text embeddings for column values ​​associated with a format structure. In certain embodiments, the data mapping component 304 learns one or more vector representations of the one or more text embeddings associated with the column names and / or column values.

[0074] The data mapping component 304 generates one or more features associated with the format structure of the heterogeneous data 314 based on one or more feature generation techniques. In an embodiment, the data mapping component 304 generates one or more features associated with the format structure for the heterogeneous data 314 based on a classifier trained on TF-IDF and / or N-gram features associated with natural language processing, where each portion of the heterogeneous data 314 is converted to a numerical format represented by a matrix. In another embodiment, the data mapping component 304 generates one or more features associated with the format structure for the heterogeneous data 314 based on SIF, where sentence embedding is calculated using word vector averaging of one or more portions of the heterogeneous data 314. In another embodiment, the data mapping component 304 generates one or more features associated with the format structure for the heterogeneous data 314 based on a universal sentence encoder, which encodes one or more portions of the heterogeneous data 314 into a dimensional vector to facilitate text classification and / or other natural language processing associated with one or more portions of the heterogeneous data 314. In another embodiment, the data mapping component 304 generates one or more features associated with a formatting structure for the heterogeneous data 314 based on a BERT embedding technique that facilitates text classification and / or other natural language processing associated with one or more portions of the heterogeneous data 314 using tokens associated with a classification task. Additionally or alternatively, the data mapping component 304 generates one or more features associated with a formatting structure for the heterogeneous data 314 based on a library of trained word embeddings and / or text classifications associated with natural language processing. In particular embodiments, the data mapping component 304 generates one or more features based on lexical ground truth data associated with one or more templates. For example, in one or more embodiments, the data mapping component 304 generates lexical ground truth data for a formatting structure based on one or more templates associated with the historical heterogeneous data.Additionally, based on the vocabulary ground truth data associated with the historical disparate data, the data mapping component 304 generates one or more features.

[0075] In one or more embodiments, the data mapping component 304 maps each portion of the heterogeneous data 314 based on one or more features to provide a formatted version of the heterogeneous data 314. In embodiments, the data mapping component 304 maps each portion of the heterogeneous data 314 based on one or more text embeddings associated with column names in the formatting structure. Further, in one or more embodiments, the data mapping component 304 maps each portion of the heterogeneous data 314 based on decision tree classifications associated with column names in the formatting structure. In particular embodiments, the data mapping component 304 calculates one or more similarity scores between one or more source column names and one or more defined target column names to facilitate mapping each portion of the heterogeneous data 314 to provide a formatted version of the heterogeneous data 314. In particular embodiments, the data mapping component 304 maps each portion of the heterogeneous data 314 based on a set of transformer encoder layers associated with a neural network. Additionally or alternatively, in particular embodiments, the data mapping component 304 maps each portion of the heterogeneous data 314 based on a text classifier associated with the neural network.

[0076] In certain embodiments, the data mapping component 304 maps source column names to target column names using one or more column values. For example, in certain embodiments, the data mapping component 304 predicts target column mappings for one or more portions of heterogeneous data using a list of column values ​​for source columns. In one example, the data mapping component 304 maps the source column name “kunnr” to the target column name “sold_to_customer_number” using a source column value of “280460-HSPL-3493664-280460.” In another example, the data mapping component 304 maps the source column name “prctr” to the target column name “profit_center_name” using a source column value of “MMS-AUTOMATIC DETECTION.” In another example, the data mapping component 304 maps the source column name “matx” to the target column name “material_number” using a source column value of “ZMPN00000000019156.” In another example, the data mapping component 304 maps the source column name "kunplz" to the target column name "sold_to_zip_code" using a source column value of "30303."

[0077] In embodiments, the artificial intelligence component 306 performs a deep learning process on the formatted version of the disparate data 314. For example, in one or more embodiments, the artificial intelligence component 306 performs a deep learning process on the formatted version of the disparate data 314 to determine one or more classifications, one or more inferences, and / or one or more insights associated with the disparate data 314. In particular embodiments, the deep learning process performed by the artificial intelligence component 306 uses regression analysis to determine one or more insights related to the disparate data 314. In particular embodiments, the deep learning process performed by the artificial intelligence component 306 uses clustering techniques to determine one or more insights associated with the disparate data 314. In one or more embodiments, the artificial intelligence component 306 performs a deep learning process to determine one or more categories and / or one or more patterns associated with the disparate data 314. In one or more embodiments, the artificial intelligence component 306 uses a recurrent neural network to map the disparate data 314 to multidimensional word embeddings in an ontology tree structure. In embodiments, the word embeddings correspond to nodes in the ontology tree structure. In one or more embodiments, the artificial intelligence component 306 uses a network of gated recurrent units of a recurrent neural network to provide one or more classifications, one or more inferences, and / or one or more insights associated with the heterogeneous data 314.

[0078] In one or more embodiments, the data optimization computer system 302 (e.g., the action component 308 of the data optimization computer system 302) receives a request 320. In an embodiment, the request 320 is a request to obtain one or more insights regarding the disparate data 314. In one or more embodiments, the request 320 includes an insight descriptor that describes one or more insight goals. In one or more embodiments, the goal is a desired data analysis result and / or target associated with the disparate data 114. In an embodiment, the insight descriptor is a word or phrase that describes the one or more insight goals. In another embodiment, the insight descriptor is an identifier that describes the one or more insight goals. In yet another embodiment, the insight descriptor is a subject that describes the one or more insight goals. However, it should be appreciated that in certain embodiments, the insight descriptor is another type of descriptor that describes the one or more insight goals. In one or more embodiments, the goal is an unclassified spend goal, a payment period optimization goal, an alternative supplier recommendation goal, and / or another insight goal. In various embodiments, the request 320 is generated by an electronic interface of a computing device. In an exemplary embodiment, the request 320 includes a request to obtain one or more insights regarding unclassified spend for one or more assets and / or services associated with the disparate data 314. Additionally, in one or more embodiments, the artificial intelligence component 306 performs a deep learning process to provide one or more insights regarding unclassified spend related to the one or more assets and / or services. In another exemplary embodiment, the request 320 includes a request to obtain one or more insights regarding payment term optimization for one or more assets and / or services associated with the disparate data 314. Furthermore, in one or more embodiments, the artificial intelligence component 306 performs a deep learning process to provide one or more insights for payment term optimization related to the one or more assets and / or services.In another exemplary embodiment, the request 320 includes a request to obtain one or more insights regarding alternative suppliers for one or more assets and / or services associated with the disparate data 314. Further, in one or more embodiments, the artificial intelligence component 306 performs a deep learning process to provide one or more insights into alternative suppliers related to the one or more assets and / or services.

[0079] In one or more embodiments, in response to the request 320, the action component 308 associates aspects of the formatted version of the disparate data 314 to provide one or more insights. In one aspect, the action component 308 determines the associated aspects of the formatted version of the disparate data 314 based on the goal and / or relationships between aspects of the formatted version of the disparate data 314. Additionally, in one or more embodiments, the action component 308 performs one or more actions based on the one or more insights. For example, in one or more embodiments, the action component 308 generates action data 322 associated with the one or more actions. In one or more embodiments, the action component 308 additionally uses a scoring model based on an iteration history of the deep learning process and / or different metrics from previous actions to determine the one or more actions. For example, in one or more embodiments, the scoring model uses weights for different metrics, different conditions, and / or different rules. In one or more embodiments, the action component 308 additionally uses location data (e.g., geographic region exceptions) to modify recommendations based on one or more rules associated with the geographic location and / or to remove false positive recommendations. In one or more embodiments, the action component 308 additionally uses contract data to modify recommendations based on one or more contract terms and / or remove false positive recommendations. In one or more embodiments, the action component 308 additionally uses cost metrics (e.g., unit cost) related to one or more assets and / or services to modify recommendations and / or remove false positive recommendations for one or more assets and / or services. In one or more embodiments, the action component 308 additionally uses risk metrics (e.g., supplier risk metrics) related to one or more assets and / or services to modify recommendations and / or remove false positive recommendations for one or more assets and / or services.In a non-limiting example, the action component 308 determines that an alternate supplier of an asset and / or service is available based on a match between part numbers in different portions of the disparate data 314. In another non-limiting example, the action component 308 determines that an alternate supplier of an asset and / or service is available based on a match between part descriptions in different portions of the disparate data 314.

[0080] In an embodiment, an action from the one or more actions includes generating a user-interactive electronic interface that renders a visual representation of the one or more insights. In another embodiment, an action from the one or more actions includes sending one or more notifications associated with the one or more insights to a computing device. In another embodiment, an action from the one or more actions includes retraining one or more portions of a recurrent neural network based on the one or more insights. In another embodiment, an action from the one or more actions includes determining one or more features associated with the one or more insights and / or predicting conditions for an asset associated with the disparate data 314 based on the one or more features associated with the one or more insights. In another embodiment, an action from the one or more actions includes predicting shipping terms for an asset associated with the disparate data 314 based on the one or more insights. In another embodiment, an action from the one or more actions includes determining a total spend for a parts commodity family associated with the disparate data 314 based on the one or more insights. In another embodiment, an action from the one or more actions includes determining one or more terms for a contract related to an asset or service associated with the disparate data 314 based on the one or more insights. In another embodiment, an action from the one or more actions includes determining, based on the one or more insights, one or more terms for a sales contract related to an asset or service associated with the disparate data 314. In another embodiment, an action from the one or more actions includes optimizing payment terms related to an asset or service associated with the disparate data 314 based on the one or more insights. In another embodiment, an action from the one or more actions includes determining, based on the one or more insights, a distribution of expenditures related to an asset or service associated with the disparate data 314.In another embodiment, an action from the one or more actions includes determining an alternative supplier for an asset or service associated with the disparate data 314 based on the one or more insights. In another embodiment, an action from the one or more actions includes determining a supplier recommendation related to an asset or service associated with the disparate data 314 based on the one or more insights. In another embodiment, an action from the one or more actions includes determining a likelihood of success for a given scenario associated with the disparate data 314 based on the one or more insights. In another embodiment, an action from the one or more actions includes providing optimal process conditions for an asset associated with the disparate data 314. For example, in another embodiment, an action from the one or more actions includes adjusting a set point and / or schedule for an asset associated with the disparate data 314. In another embodiment, an action from the one or more actions includes one or more corrective actions to take on an asset associated with the disparate data 314. In another embodiment, an action from the one or more actions includes providing optimal maintenance options for an asset associated with the disparate data 314. In another embodiment, an action from the one or more actions includes an action associated with application services layer 225, application layer 230, and / or core services layer 235. In particular embodiments, data mapping component 304 updates one or more features based on quality scores associated with one or more insights. Additionally or alternatively, in particular embodiments, data mapping component 304 updates one or more features based on user feedback data associated with one or more insights.

[0081] 4 illustrates a system 300′ that provides an exemplary environment in accordance with one or more described features of one or more embodiments of the present disclosure. In an embodiment, the system 300′ corresponds to an alternative embodiment of the system 300 illustrated in FIG. 3 . According to an embodiment, the system 300′ includes a data-optimized computer system 302, one or more data sources 316, a data lake 318, and / or a computing device 402. In one or more embodiments, the data-optimized computer system 302 communicates with the one or more data sources 316 and / or the computing device 402 over the network 110. The computing device 402 is a mobile computing device, a smartphone, a tablet computer, a mobile computer, a desktop computer, a laptop computer, a workstation computer, a wearable device, a virtual reality device, an augmented reality device, or another type of computing device that is remote from the data-optimized computer system 302.

[0082] In one or more embodiments, the action component 308 communicates the action data 322 to the computing device 402. For example, in one or more embodiments, the action data 322 includes one or more visual elements for a visual display (e.g., a user-interactive electronic interface) of the computing device 402 that renders a visual representation of the one or more insights. In particular embodiments, the visual display of the computing device 402 displays one or more graphical elements (e.g., one or more insights) associated with the action data 322. In particular embodiments, the visual display of the computing device 402 provides a graphical user interface to facilitate managing data usage associated with one or more assets associated with the disparate data 314, costs associated with one or more assets associated with the disparate data 314, asset planning associated with one or more assets associated with the disparate data 314, asset services associated with one or more assets associated with the disparate data 314, asset operations associated with one or more assets associated with the disparate data 314, and / or one or more other aspects of the one or more assets associated with the disparate data 314. In particular embodiments, the visual display of computing device 402 provides a graphical user interface to facilitate predicting shipping terms for one or more assets associated with disparate data 314. In particular embodiments, the visual display of computing device 402 facilitates predicting total expenditures for one or more assets associated with disparate data 314. In another example, in one or more embodiments, action data 322 includes one or more notifications associated with one or more insights. In one or more embodiments, action data 322 enables a user associated with computing device 402 to make a decision and / or perform one or more actions regarding the one or more insights.

[0083] 5 illustrates a system 500 according to one or more embodiments of the present disclosure. System 500 includes a computing device 402. In one or more embodiments, computing device 402 uses mobile computing, augmented reality, cloud-based computing, IoT technology, and / or one or more other technologies to provide video, audio, real-time data, graphical data, one or more communications, one or more messages, one or more notifications, one or more documents, one or more work procedures, industrial asset tag details, and / or other media data associated with one or more insights. Computing device 402 includes mechanical, electrical, hardware, and / or software components to facilitate obtaining one or more insights associated with disparate data 314. In the embodiment illustrated in FIG. 5, computing device 402 includes a visual display 504, one or more speakers 506, one or more cameras 508, one or more microphones 510, a global positioning system (GPS) device 512, a gyroscope 514, one or more wireless communication devices 516, and / or a power source 518.

[0084] In embodiments, the visual display 504 is a display that facilitates presentation and / or interaction with one or more portions of the action data 322. In one or more embodiments, the computing device 402 displays an electronic interface (e.g., a graphical user interface) associated with the data analytics platform. In one or more embodiments, the visual display 504 is a visual display that renders one or more interactive media elements via a set of pixels. The one or more speakers 506 include one or more integrated speakers that emit audio. The one or more cameras 508 include one or more cameras that use autofocus and / or image stabilization for photo capture and / or real-time video. The one or more microphones 510 include one or more digital microphones that capture audio data using active noise cancellation. The GPS device 512 provides a geographic location to the computing device 402. The gyroscope 514 determines the orientation of the computing device 402. The one or more wireless communication devices 516 include one or more hardware components for wireless communication via one or more wireless networking technologies and / or one or more short-wavelength wireless technologies. Power source 518 may be a power source and / or a rechargeable battery, for example, that powers visual display 504, one or more speakers 506, one or more cameras 508, one or more microphones 510, GPS device 512, gyroscope 514, and / or one or more wireless communication devices 516. In certain embodiments, data associated with one or more insights is presented via visual display 504 and / or one or more speakers 506.

[0085] 6 illustrates a system 600 in accordance with one or more described features of one or more embodiments of the present disclosure. In an embodiment, the system 600 includes uncategorized purchase record data 602. For example, in an embodiment, the uncategorized purchase record data 602 corresponds to at least a portion of the disparate data 314 obtained from one or more data sources 316. It should be understood that in particular embodiments, the uncategorized purchase record data 602 corresponds to other uncategorized data, such as other uncategorized record data, uncategorized asset data, uncategorized industry data, etc. In one example, the uncategorized purchase record data 602 includes a data field 604 associated with supplier information, a data field 606 associated with part (e.g., asset) information, a data field 608 associated with part family codes (PFCs), and / or a data field 610 associated with spending. However, it should be understood that in particular embodiments, the uncategorized purchase record data 602 (e.g., data fields of the uncategorized purchase record data) may be associated with other information related to uncategorized spend, payment term optimization, alternative supplier recommendations, and / or other insights goals. For example, in certain embodiments, the data fields 604 may additionally or alternatively include one or more data fields associated with a purchase order number, an invoice number, a supplier identifier, a payment term, an invoice amount, a supplier hierarchy level identifier, a purchase order line item, a purchase order balance, a purchase order term, a part number, a product commodity family, a part description, and / or other information. In embodiments, the data mapping component 304 aggregates the uncategorized purchase record data 602 to generate aggregated total spend data. For example, in embodiments, the data mapping component 304 aggregates data fields 604 associated with supplier information, data fields 606 associated with part (e.g., asset) information, data fields 608 associated with PFCs, and / or data fields 610 associated with spending into a total spend for each supplier and each PFC. In one or more embodiments, the action component 308 determines the PFC with the highest spend. For example, as shown in FIG. 6 , the PFC for the top-level spending supplier S1 is C01.In one or more embodiments, the data mapping component 304 and / or the artificial intelligence component 306 uses a data mapping table 614 that maps PFCs to supplier commodity offices to determine classification data 616 for the aggregate total spend data 612. For example, in one or more embodiments, the data mapping table 614 is configured to perform a mapping between a data field (e.g., a PFC) and a particular classification to determine the classification data 616 for the aggregate total spend data 612. In one or more embodiments, the aggregate total spend data 612 is formatted as a vector of data or a matrix of data, and the data mapping table 614 is configured to rescale the dimensions of the aggregate total spend data 612 to provide different data dimensions.

[0086] FIG. 7 illustrates a machine learning model 700 according to one or more described features of one or more embodiments of the present disclosure. In an embodiment, the machine learning model 700 is a recurrent neural network. In another embodiment, the machine learning model 700 is a convolutional neural network. In another embodiment, the machine learning model 700 is a deep learning network. However, it should be appreciated that in certain embodiments, the machine learning model 700 is another type of artificial neural network. In one or more embodiments, an input sequence 702 is provided as input to the machine learning model 700. In various embodiments, the input sequence 702 includes a set of data elements associated with the heterogeneous data 314. In one or more embodiments, the data mapping component 304 uses the machine learning model 700 (e.g., a recurrent neural network) to map the input sequence 702 to multidimensional word embeddings 704. For example, in one or more embodiments, each portion of the input sequence 702 is converted to a respective multidimensional word embedding 704. In one or more embodiments, each word associated with the input sequence 702 is mapped to a respective vector associated with the multidimensional word embedding 704. In embodiments, a multidimensional word embedding of multidimensional word embeddings 704 is a vector of data or a matrix of data to facilitate one or more deep learning processes, such as natural language processing. In one or more embodiments, artificial intelligence component 306 provides multidimensional word embeddings 704 to a network of gated recurrent units 706. In embodiments, a gated-recurrent unit (GRU) from the network of gated recurrent units 706 is a gating mechanism having an update gate and / or a reset gate that determines the data that passes as the output of the gated recurrent unit.For example, in embodiments, the update gate determines the amount of data that is forwarded along the network of gated recurrent units 706 (e.g., how much previous data is provided from a previous state of the network of gated recurrent units 706 to a next state of the network of gated recurrent units 706), and the reset gate determines the amount of data that is withheld from being forwarded along the network of gated recurrent units 706 (e.g., how much previous data is withheld from a next state of the network of gated recurrent units 706). In one or more embodiments, the output data from the network of gated recurrent units 706 undergoes a concatenation process that combines data from each gated recurrent unit in the network of gated recurrent units 706. In particular embodiments, the concatenated output 708 of the network of gated recurrent units 706 is processed by a first tightly coupled layer 710 (e.g., tightly coupled layer 32) and / or a tightly coupled layer 712 (e.g., tightly coupled layer 16), which changes the dimensionality of the concatenated output of the network of gated recurrent units 706. Further, based on the concatenated outputs of the network of gated recurrent units 706, the densely-coupled layer 710, and / or the densely-coupled layer 716, the machine learning model 700 provides a prediction 714. In one or more embodiments, the prediction 714 relates to one or more insights regarding the input sequence 702 (e.g., regarding a set of data elements associated with the heterogeneous data 314). For example, in one or more embodiments, the prediction 714 includes one or more classifications regarding the input sequence 702 (e.g., regarding a set of data elements associated with the heterogeneous data 314). In embodiments, the input sequence 702 includes one or more words from the heterogeneous data 314 that are converted into respective multidimensional word embeddings 704 associated with respective vectors of data. Each GRU from the network of gated recurrent units 706 processes each multidimensional word embedding 704 to provide a concatenated output 708 that combines the outputs from each GRU from the network of gated recurrent units 706.In particular embodiments, the dimensionality of the concatenated output 708 is changed via the first densely connected layer 710 and / or the densely connected layer 712 to provide a predicted classification (e.g., prediction 714) for one or more words from the heterogeneous data 314.

[0087] FIG. 8 illustrates a system 800 according to one or more embodiments of the present disclosure. System 800, for example, provides a mapping model architecture. Furthermore, system 800 illustrates one or more embodiments relating to data mapping component 304. In one or more embodiments, heterogeneous data 314 is processed by column name model process 802 and / or column value model process 804. Column name model process 802 is used to provide one or more column name features, classifications, and / or mapping recommendations associated with the format structure of one or more portions of heterogeneous data 314. In an embodiment, column name model process 802 includes feature generation 806. Feature generation 806 generates one or more column name features for heterogeneous data 314. For example, feature generation 806 performs feature generation based on column names to provide input data (e.g., one or more column name features) for classification model 808. In particular embodiments, feature generation 806 generates one or more column name features for the heterogeneous data 314 based on a TF-IDF technique, a SIF technique, a universal sentence encoder technique, a BERT embedding technique, and / or another feature generation technique. In particular embodiments, feature generation 806 generates one or more column name features for the heterogeneous data 314 based on a library of learned word embeddings and / or text classification associated with natural language processing. The classification model 808 is, for example, a trained classification model and provides one or more inferences associated with the heterogeneous data 314 and / or one or more column name features of the heterogeneous data 314. In embodiments, the classification model 808 is a tree-based classification model. For example, in one or more embodiments, the classification model 808 is a hierarchical classification model and includes at least a first level associated with predicting dataset categories and a second level associated with predicting corresponding column names using the predicted dataset categories as features. Further, in embodiments, the classification model 808 generates at least a portion of one or more mapping recommendations 810. In particular embodiments, the column name model process 802 includes training 812 to train the classification model 808 .In one or more embodiments, training 812 trains classification model 808 using one or more column name features generated based on training data 814. Training data 814 includes, for example, lexical ground truth data for format structures generated based on one or more templates associated with historical column name features. In particular embodiments, training data 814 includes predetermined target data associated with column name features.

[0088] Column value model processing 804 may additionally or alternatively be used to provide one or more column value features, classifications, and / or mapping recommendations associated with the format structure of one or more portions of heterogeneous data 314. In embodiments, column value model processing 804 includes feature generation 816. Feature generation 816 generates one or more column value features for heterogeneous data 314. For example, feature generation 816 performs feature generation based on column values ​​to provide input data (e.g., one or more column value features) for classification model 818. Classification model 818 may be, for example, a trained classification model, and may provide one or more inferences associated with heterogeneous data 314 and / or one or more column value features for heterogeneous data 314. In embodiments, classification model 818 is a transformer-based classification model. For example, in one or more embodiments, classification model 818 is a neural network and may include a set of transformer encoder layers, a set of hidden layers, a set of attention layers, and / or a densely connected layer. Further, in embodiments, classification model 818 generates at least a portion of one or more mapping recommendations 810. For example, in embodiments, the classification model 818 provides a predicted target column mapping based on a set of column values ​​associated with the heterogeneous data 314. In particular embodiments, the column value model processing 804 includes training 820, which trains the classification model 818. In one or more embodiments, the training 820 trains the classification model 818 using one or more column value features generated based on training data 822. The training data 822 includes, for example, lexical ground truth data of format structures generated based on one or more templates associated with the historical column value features. In particular embodiments, the one or more mapping recommendations 810 are ranked based on their respective confidence scores to provide the top-N mapping recommendations. In particular embodiments, the one or more mapping recommendations 810 are associated with a probability distribution of the mapping recommendations. In particular embodiments, the one or more mapping recommendations 810 are accepted by the data optimization computer system 302 and / or via user feedback associated with the computing device 402.In certain embodiments, classification model 808 and / or classification model 818 are retrained based on one or more mapping recommendations 810. For example, in certain embodiments, classification model 808 and / or classification model 818 are retrained based on one or more mapping recommendations 810 that have been accepted by data optimization computer system 302. Additionally or alternatively, in certain embodiments, classification model 808 and / or classification model 818 are retrained based on user feedback associated with computing device 402.

[0089] FIG. 9 illustrates a system 900 according to one or more embodiments of the present disclosure. The system 900 may, for example, provide a mapping model architecture. In one or more embodiments, the system 900 may provide a column name model architecture related to the classification model 808. Additionally, the system 900 may illustrate one or more embodiments related to the data mapping component 304. The system 900 may include a ground truth model 902, a supervised model 904, a text similarity supervised model 906, and / or a feature similarity unsupervised model 908. In one or more embodiments, a source template 910 and / or a target template 912 may be provided as input to the ground truth model 902. The source template 910 may, for example, be a template of a source format structure for one or more portions of the heterogeneous data 314 associated with one or more data sources 316. The target template 912 may, for example, be a template of a target format structure for storing one or more portions of the heterogeneous data 314 in the data lake 318. For example, in one or more embodiments, source template 910 is associated with a set of source column names, and target template 912 is associated with a set of target column names. In particular embodiments, source data 914 and / or target data 916 are additionally or alternatively provided as inputs to ground truth model 902. For example, in embodiments, source data 914 is source data stored in source template 910, and target data 916 is historical target data stored in target template 912. In one or more embodiments, one or more portions of disparate data 314 correspond to source data 914.

[0090] In one or more embodiments, the ground truth model 902 uses the source template 910, the target template 912, the source data 914, and / or the target data 916 to generate vocabulary (e.g., vocabulary ground truth data) and / or features (e.g., feature ground truth data) for data field mappings related to the format structure. In one or more embodiments, the supervised model 904 is used to predict mappings for one or more data field mappings that do not meet a certain confidence threshold. For example, in one or more embodiments, the supervised model 904 predicts mappings of target data fields for the target format structure to source data fields for the source format structure. In certain embodiments, the supervised model 904 is retrained based on at least a portion of the target data 916. In certain embodiments, at least a portion of the target data 916 is provided via the computing device 402.

[0091] In one or more embodiments, a text similarity supervised model 906 is used to predict mappings for one or more data field mappings that do not meet a particular confidence threshold. For example, in particular embodiments, a text similarity supervised model 906 is used to predict mappings for one or more data field mappings that do not meet a particular confidence threshold following processing by a supervised model. In one or more embodiments, the text similarity supervised model 906 determines text similarity between data field names and / or data field descriptions of the target format structure and the source format structure. In an exemplary embodiment, the target data field name is "BRGEW" and the data field description is "weight." Thus, in one example, the text similarity supervised model 906 determines that the data field description "weight" corresponds to "unit weight of ingredient." In another example, the text similarity supervised model 906 determines that the data field description "weight" corresponds to "weight of ingredient." In another example, the text similarity supervised model 904 determines that the data field description "weight" corresponds to the "weight" data field description for a particular target format structure.

[0092] In one or more embodiments, the feature similarity supervised model 908 is used to predict mappings for one or more data field mappings that do not meet a particular confidence threshold. For example, in certain embodiments, the feature similarity supervised model 908 is used to predict mappings for one or more data field mappings that do not meet a particular confidence threshold following processing by a supervised model and / or a text similarity supervised model. In one or more embodiments, the feature similarity supervised model 908 is configured to analyze and / or identify data characteristics related to the source data 914. Additionally or alternatively, in one or more embodiments, the feature similarity supervised model 908 determines feature matrix similarities between the source data 914 and the target data 916. In one or more embodiments, the feature similarity supervised model 908 provides mapping recommendations 918. The mapping recommendations 918 are, for example, at least a portion of one or more mapping recommendations 810. In embodiments, the mapping recommendations 918 include one or more mapping recommendations for the source data 914 (e.g., mapping recommendations for one or more portions of the heterogeneous data 314). In another embodiment, the mapping recommendation 918 includes predicted column name data fields of a format structure for the source data 914 (e.g., one or more portions of the heterogeneous data 314). In particular embodiments, the mapping recommendation 918 provides a formatted version of the source data 914 (e.g., one or more portions of the heterogeneous data 314). In particular embodiments, the mapping recommendation 918 categorizes one or more portions of the source data 914 as respective specified column name labels.

[0093] In one or more embodiments, the ground truth model 902 maps a context vocabulary generated from historical data. In particular embodiments, the historical data is associated with data objects such as a “customer master,” “vendor master,” “material master,” “bill of materials,” “routing,” “purchase information record,” and / or other data objects. In one or more embodiments, to enhance the ground truth model 902, valid and / or invalid tokens are defined using historical mapping information and / or by analyzing trained model results. In one or more embodiments, the valid tokens are used to recommend possible similar mappings for fields. In one or more embodiments, the invalid tokens are used to eliminate model recommendations that exhibit the same or similar data characteristics. In one or more embodiments, the eliminated model recommendations are also deemed irrelevant. The supervised model 904 is configured to perform the mapping based on field names. In one or more embodiments, the supervised model 904 uses one or more natural language processing techniques to learn one or more patterns associated with field names. The text similarity supervised model 906 is configured to perform the mapping based on field descriptions. In one or more embodiments, the text similarity supervised model 906 performs similarity checks between field descriptions of a system, database, and / or data model. For example, in one or more embodiments, the text similarity supervised model 906 is used to identify mapping similarities between field descriptions of a system, database, and / or data model. In particular embodiments, the text similarity supervised model 906 runs two or more text similarity models to identify mapping similarities between field descriptions for a system, database, and / or data model. In particular embodiments, the best recommendation associated with the two or more text similarity models is selected.

[0094] The feature similarity unsupervised model 908 is configured to perform mappings based on data features. In one or more embodiments, the feature similarity unsupervised model 908 analyzes data to learn mappings between systems, databases, and / or data models. In one or more embodiments, the feature similarity unsupervised model 908 compares features associated with data using one or more similarity algorithms. In one or more embodiments, the feature similarity unsupervised model 908 separates features based on data type, such as numeric features, text features, date features, and / or another data type. Examples of numeric features include, but are not limited to, mean, median, standard deviation, skewness, and / or another numeric feature. Examples of text features include statistics based on spaces, numbers, letters, parentheses, special characters, and / or other features. In one or more embodiments, the feature similarity unsupervised model 908 determines custom features by searching for one or more specific patterns in the data and / or identifying keywords for one or more of the data fields. In one or more embodiments, the unsupervised feature similarity model 908 clusters data fields into unique categories to reduce the size of the search space for the data. Thus, in one or more embodiments, the amount of time and / or computing resources required to perform the feature comparison process is reduced.

[0095] In embodiments, source template 910 is a first template that includes a first template format configured with a first dimension associated with a first set of columns and / or column names. Additionally, target template 912 is a second template that includes a second template format configured with a second dimension associated with a second set of columns and / or column names. In one or more embodiments, source data 914 includes asset data stored in source template 910 (e.g., asset data associated with edge devices 161a-n), and target data 916 is historical asset data stored in target template 912. In one or more embodiments, ground truth model 902 generates vocabulary (e.g., vocabulary ground truth data) and / or features (e.g., feature ground truth data) for the asset data associated with source data 914 and the historical asset data associated with target data 916. The vocabulary and / or features for the asset data associated with the source data 914 and / or the historical asset data associated with the target data 916 include, for example, asset names, asset statuses, real-time asset values, target values, field status values, importance indicators, one or more asset rules, one or more asset requirements, text embeddings, etc. Additionally, in one or more embodiments, the supervised model 904 predicts a mapping of target data fields for the target template 912 to source data fields for the source template 910. In one or more embodiments, the text similarity supervised model 906 determines text similarities between data field names and / or data field descriptions of the target format structure 910 and the source format structure 912. For example, in embodiments, the text similarity supervised model 906 determines that the data field description "field status" in the source format structure 912 corresponds to "asset status" in the target format structure 910.In one or more embodiments, the feature similarity supervised model 908 is configured to analyze and / or identify data characteristics associated with the asset data associated with the source data 914 and / or the historical asset data associated with the target data 916. In one or more embodiments, the mapping recommendation 918 provides predicted column name data fields of a format structure in the target template 912 for the source data 914 associated with the asset data.

[0096] FIG. 10 illustrates a system 1000 in accordance with one or more embodiments of the present disclosure. In an embodiment, the system 1000 corresponds to a Transformer-based classification model. In one or more embodiments, the system 1000 provides a column-value model architecture related to the classification model 818. Additionally, the system 1000 illustrates one or more embodiments related to the data mapping component 304. In one or more embodiments, input data 1002 is provided to a set of Transformer layers 1004a-n of the system 1000. The input data 1002 corresponds to one or more portions of the heterogeneous data 314. In one or more embodiments, the input data 1002 includes, for example, one or more column values ​​associated with the heterogeneous data 314. In one or more embodiments, the set of Transformer layers 1004a-n learns one or more relationships and / or one or more features between the input data 1002. Each Transformer layer from the set of Transformer layers 1004a-n includes a respective weight and / or a respective bias to facilitate learning one or more relationships and / or one or more features between the input data 1002. For example, in one or more embodiments, the set of Transformer layers 1004a-n learns one or more relationships and / or one or more features between characters included in the input data 1002. In an embodiment, Transformer layer 1004a provides data 1008 associated with a first learned relationship and / or feature associated with the input data 1002. Additionally, Transformer layer 1004b learns one or more relationships and / or one or more features associated with the data 1008 to provide data 1010 associated with a second learned relationship and / or feature. In this embodiment, Transformer layer 1004n also learns one or more relationships and / or one or more features to provide a Transformer layer output 1012 associated with n learned relationships and / or features, where n is an integer. The transformer layer output 1012 is provided as an input to the classifier 1006, which uses the transformer layer output 1012 to provide a mapping recommendation 1014. The mapping recommendation 1014 is, for example, at least a portion of one or more of the mapping recommendations 810.In an embodiment, mapping recommendation 1014 includes one or more mapping recommendations for input data 1002 (e.g., mapping recommendations for one or more portions of heterogeneous data 314). In another embodiment, mapping recommendation 1014 includes predicted column name data fields of a format structure for input data 1002 (e.g., one or more portions of heterogeneous data 314). In particular embodiments, mapping recommendation 1014 provides a formatted version of input data 1002 (e.g., one or more portions of heterogeneous data 314). In particular embodiments, mapping recommendation 1014 classifies one or more portions of input data 1002 as respective defined column name labels.

[0097] FIG. 11 illustrates a system 1100 according to one or more embodiments of the present disclosure. In an embodiment, the system 1100 corresponds to a neural network architecture related to the classification model 818. Additionally, the system 1000 illustrates one or more embodiments related to the data mapping component 304. In one or more embodiments, input string values ​​1102 undergo character-level embeddings 1104. The input string values ​​1102 correspond, for example, to at least a portion of the heterogeneous data 314. Additionally, in one or more embodiments, the output of the character-level embeddings 1104 is provided to a Transformer 1106, which provides the Transformer layer output to a Classifier 1108. In particular embodiments, the Transformer 1106 corresponds to the set of Transformer layers 1004a-n, which corresponds to the classifier 1006. The classifier 1108 provides a mapping recommendation 1110. The mapping recommendation 1110 is, for example, at least a portion of one or more mapping recommendations 810. In an embodiment, mapping recommendation 1110 includes one or more mapping recommendations for input column values ​​1102. In another embodiment, mapping recommendation 1110 includes a predicted column name data field of a format structure for input column values ​​1102. In particular embodiments, mapping recommendation 1110 provides a formatted version of input column values ​​1102. In particular embodiments, mapping recommendation 1110 categorizes input column values ​​1102 using defined column name labels.

[0098] 12 illustrates a method 1200 for performing optimization related to enterprise performance management, according to one or more embodiments described herein. Method 1200 may be associated with, for example, the data optimization computer system 302. For example, in one or more embodiments, method 1200 is executed on a device (e.g., the data optimization computer system 302) having one or more processors and memory. In one or more embodiments, method 1200 begins at block 1202 with receiving (e.g., by the data mapping component 304) a request to obtain one or more insights regarding a formatted version of disparate data associated with one or more data sources, the request including an insight descriptor that describes one or more insight goals (block 1202). The request to obtain the one or more insights may result in one or more technology improvements, such as, but not limited to, facilitating interaction with a computing device, enhancing the functionality of the computing device, and / or improving the accuracy of data provided to the computing device.

[0099] At block 1204, it is determined whether the request is processed. If no, block 1204 is repeated to determine whether the request is processed. If yes, method 1200 proceeds to block 1206. In response to the request, block 1206 associates aspects of the formatted versions of the disparate data (e.g., by the artificial intelligence component 306) to provide one or more insights, the associated aspects being determined by the goal and the relationships between the aspects of the formatted versions of the disparate data. Associating aspects of the formatted versions of the disparate data may result in one or more technical improvements, such as, but not limited to, extending the functionality of the computing device and / or improving the accuracy of data provided to the computing device. In one or more embodiments, associating aspects of the formatted versions of the disparate data includes correlating aspects of the formatted versions of the disparate data to provide one or more insights. In one or more embodiments, correlating aspects of the formatted versions of the heterogeneous data includes using machine learning associated with a machine learning model, a ground truth model, a supervised model, a text similarity supervised model, a feature similarity unsupervised model, a column name modeling, a column value modeling, a classifier, and / or another type of machine learning technique.

[0100] Method 1200 also includes block 1208, which performs one or more actions (e.g., by action component 308) based on the one or more insights. Performing the one or more actions results in one or more technology improvements, such as, but not limited to, providing a different experience on a computing device and / or providing a visual indicator via a computing device. In one or more embodiments, the one or more actions include generating a user-interactive electronic interface that renders a visual representation of the one or more insights. In one or more embodiments, the one or more actions include sending, to the computing device, one or more notifications associated with the one or more insights. In one or more embodiments, the one or more actions include predicting shipping terms for assets associated with the disparate data based on the one or more insights. In one or more embodiments, the one or more actions include determining part commodity families for uncategorized purchase record data associated with the disparate data based on the one or more insights. In one or more embodiments, the one or more actions include determining a total spend for the part commodity families based on the one or more insights.

[0101] In one or more embodiments, method 1200 further includes aggregating the disparate data from one or more data sources. Aggregating the disparate data from one or more data sources may result in one or more technology improvements, such as, but not limited to, expanding the capabilities of the computing device and / or improving the accuracy of the data provided to the computing device. In one or more embodiments, aggregating the disparate data includes storing the disparate data in a single data lake and / or updating the data in the single data lake at one or more predetermined intervals.

[0102] In one or more embodiments, method 1200 further includes formatting one or more portions of the disparate data, where formatting provides a formatted version of the disparate data associated with the defined format. Formatting one or more portions of the disparate data may also result in one or more technology improvements, such as, but not limited to, expanding the functionality of the computing device and / or improving the accuracy of the data provided to the computing device. In one or more embodiments, method 1200 further includes determining one or more mapping recommendations for the formatted version of the disparate data. In one or more embodiments, formatting one or more portions of the disparate data includes identifying one or more distinct data fields in the disparate data from one or more data sources, where the distinct data fields describe corresponding subject matter. Additionally, in one or more embodiments, formatting one or more portions of the disparate data includes determining one or more incomplete data fields from the one or more data sources, where the one or more incomplete data fields correspond to the identified one or more distinct data fields. In one or more embodiments, formatting one or more portions of the disparate data further includes adding data from the identified data fields to the incomplete data fields in accordance with a determination that the one or more determined incomplete data fields from the one or more data sources correspond to the identified one or more different data fields. In one or more embodiments, formatting one or more portions of the disparate data includes organizing the formatted version of the disparate data based on an ontology tree structure that captures relationships between different data within the disparate data. In one or more embodiments, method 1200 further includes comparing the different data sources based on the ontology tree structure. In one or more embodiments, associating aspects of the formatted versions of the disparate data includes traversing the ontology tree structure, wherein traversing associates aspects of the disparate data.The ontology tree structure provides one or more technical improvements, such as, but not limited to, extending the functionality of a computing device, improving the accuracy of data provided to the computing device, and / or improving the efficiency of the computing device.

[0103] In one or more embodiments, method 1200 further includes performing a deep learning process on the formatted version of the heterogeneous data to provide one or more insights associated with the heterogeneous data. In one or more embodiments, performing the deep learning process includes determining one or more classifications for the formatted version of the heterogeneous data to provide the one or more insights. In one or more embodiments, performing the deep learning process includes mapping the heterogeneous data to multidimensional word embeddings using a recurrent neural network. In one or more embodiments, performing the deep learning process includes providing the one or more insights using a network of gated recurrent units of the recurrent neural network. Performing the deep learning process results in one or more technical improvements, such as, but not limited to, enhancing the functionality of the computing device and / or improving the accuracy of the data provided to the computing device. In one or more embodiments, method 1200 further includes retraining one or more portions of the recurrent neural network based on the one or more insights. Retraining one or more portions of the recurrent neural network results in one or more technical improvements, such as, but not limited to, improving the accuracy of the recurrent neural network. In one or more embodiments, method 1200 further includes determining one or more actions using a scoring model based on different metrics from the iteration history of the deep learning process. Using the scoring model results in one or more technical improvements, such as, but not limited to, extending the capabilities of the computing device and / or improving the accuracy of data provided to the computing device.

[0104] 13 illustrates a method 1300 for performing optimization related to enterprise performance management, according to one or more embodiments described herein. Method 1300 may be associated with, for example, data optimization computer system 302. For example, in one or more embodiments, method 1300 is performed on a device (e.g., data optimization computer system 302) having one or more processors and memory. In one or more embodiments, method 1300 begins at block 1302, which generates (e.g., by data mapping component 304) one or more features associated with a format structure for heterogeneous data associated with one or more data sources. In one or more embodiments, generating the one or more features includes generating one or more text embeddings associated with column names for the format structure. Generating the one or more features may result in one or more technical improvements, such as, but not limited to, enhancing the functionality of a computing device and / or improving the accuracy of data provided to the computing device.

[0105] At block 1304, each portion of the heterogeneous data is mapped (e.g., by the data mapping component 304) based on the one or more features to provide a formatted version of the heterogeneous data. In one or more embodiments, the mapping includes mapping each portion of the heterogeneous data based on one or more text embeddings associated with column names for the formatting structure. In one or more embodiments, the mapping additionally or alternatively includes mapping each portion of the heterogeneous data based on decision tree classifications associated with column names for the formatting structure. In one or more embodiments, the mapping additionally or alternatively includes learning one or more vector representations of the one or more text embeddings associated with the column names. In one or more embodiments, the mapping additionally or alternatively includes calculating one or more similarity scores between the one or more source column names and the one or more defined target column names. In one or more embodiments, the mapping additionally or alternatively includes generating one or more text embeddings associated with column values ​​of the formatting structure. In one or more embodiments, the mapping additionally or alternatively includes mapping each portion of the heterogeneous data based on a set of Transformer encoder layers associated with a neural network. In one or more embodiments, the mapping additionally or alternatively includes mapping each portion of the disparate data based on a text classifier associated with a neural network, wherein the mapping of each portion of the disparate data results in one or more technical improvements, such as, but not limited to, enhancing the functionality of the computing device and / or improving the accuracy of the data provided to the computing device.

[0106] At 1306, a request to obtain one or more insights about the formatted version of the disparate data is received (e.g., by the data mapping component 304), the request including an insight descriptor that describes a goal of the one or more insights (block 1302). The request to obtain the one or more insights may result in one or more technology improvements, such as, but not limited to, facilitating interaction with the computing device, extending the functionality of the computing device, and / or improving the accuracy of data provided to the computing device.

[0107] At block 1308, it is determined whether the request is processed. If no, block 1308 is repeated to determine whether the request is processed. If yes, method 1300 proceeds to block 1310. In response to the request, block 1310 associates aspects of the formatted versions of the disparate data (e.g., by the artificial intelligence component 306) to provide one or more insights, the associated aspects being determined by the goal and the relationships between the aspects of the formatted versions of the disparate data. Associating aspects of the formatted versions of the disparate data may result in one or more technical improvements, such as, but not limited to, extending the functionality of the computing device and / or improving the accuracy of data provided to the computing device. In one or more embodiments, associating aspects of the formatted versions of the disparate data includes correlating aspects of the formatted versions of the disparate data to provide one or more insights. In one or more embodiments, correlating aspects of the formatted versions of the heterogeneous data includes using machine learning associated with a machine learning model, a ground truth model, a supervised model, a text similarity supervised model, a feature similarity unsupervised model, a column name modeling, a column value modeling, a classifier, and / or another type of machine learning technique.

[0108] Method 1300 also includes block 1312, which performs one or more actions (e.g., by action component 308) based on the one or more insights. Performing the one or more actions results in one or more technology improvements, such as, but not limited to, providing a different experience on a computing device and / or providing a visual indicator via a computing device. In one or more embodiments, the one or more actions include generating a user-interactive electronic interface that renders a visual representation of the one or more insights. In one or more embodiments, the one or more actions include sending, to a computing device, one or more notifications associated with the one or more insights. In one or more embodiments, the one or more actions include predicting shipping terms for assets associated with the disparate data based on the one or more insights. In one or more embodiments, the one or more actions include determining part commodity families for uncategorized purchase record data associated with the disparate data based on the one or more insights. In one or more embodiments, the one or more actions include determining a total spend for the part commodity families based on the one or more insights.

[0109] In one or more embodiments, method 1300 further includes providing one or more mapping recommendations for the formatted version of the disparate data based on the one or more insights. Additionally or alternatively, in one or more embodiments, method 1300 further includes updating one or more features based on the one or more mapping recommendations. Providing one or more mapping recommendations and / or updating one or more features may result in one or more technical improvements, such as, but not limited to, enhancing the functionality of the computing device and / or improving the accuracy of the data provided to the computing device.

[0110] In one or more embodiments, method 1300 further includes generating lexical ground truth data for the format structure based on one or more templates associated with the historical heterogeneous data. Further, in one or more embodiments, generating the one or more features includes generating the one or more features based on the lexical ground truth data associated with the one or more templates. Generating the lexical ground truth data may result in one or more technical improvements, such as, but not limited to, enhancing the functionality of the computing device and / or improving the accuracy of data provided to the computing device.

[0111] In one or more embodiments, method 1300 further includes updating the one or more features based on a quality score associated with the one or more insights. Additionally or alternatively, in one or more embodiments, method 1300 further includes updating the one or more features based on user feedback data associated with the one or more insights. Updating the one or more features may result in one or more technical improvements, such as, but not limited to, enhancing the functionality of the computing device and / or improving the accuracy of data provided to the computing device.

[0112] In one or more embodiments, method 1300 further includes aggregating the disparate data from one or more data sources. Aggregating the disparate data from one or more data sources may result in one or more technology improvements, such as, but not limited to, expanding the capabilities of the computing device and / or improving the accuracy of the data provided to the computing device. In one or more embodiments, aggregating the disparate data includes storing the disparate data in a single data lake and / or updating the data in the single data lake at one or more predetermined intervals.

[0113] In one or more embodiments, method 1300 further includes formatting one or more portions of the disparate data, where formatting provides a formatted version of the disparate data associated with the defined format. Formatting one or more portions of the disparate data may also result in one or more technical improvements, such as, but not limited to, expanding the functionality of the computing device and / or improving the accuracy of the data provided to the computing device. In one or more embodiments, method 1300 further includes determining one or more mapping recommendations for the formatted version of the disparate data. In one or more embodiments, formatting one or more portions of the disparate data includes identifying one or more distinct data fields in the disparate data from one or more data sources, where the distinct data fields describe corresponding subject matter. Additionally, in one or more embodiments, formatting one or more portions of the disparate data includes determining one or more incomplete data fields from the one or more data sources, where the one or more incomplete data fields correspond to the identified one or more distinct data fields. In one or more embodiments, formatting one or more portions of the disparate data further includes adding data from the identified data fields to the incomplete data fields in accordance with a determination that the one or more determined incomplete data fields from the one or more data sources correspond to the identified one or more different data fields. In one or more embodiments, formatting one or more portions of the disparate data includes organizing the formatted version of the disparate data based on an ontology tree structure that captures relationships between different data within the disparate data. In one or more embodiments, method 1300 further includes comparing the different data sources based on the ontology tree structure. In one or more embodiments, associating aspects of the formatted versions of the disparate data includes traversing the ontology tree structure, wherein traversing associates aspects of the disparate data.The ontology tree structure provides one or more technical improvements, such as, but not limited to, extending the functionality of a computing device, improving the accuracy of data provided to the computing device, and / or improving the efficiency of the computing device.

[0114] In one or more embodiments, method 1300 further includes performing a deep learning process on the formatted version of the heterogeneous data to provide one or more insights associated with the heterogeneous data. In one or more embodiments, performing the deep learning process includes determining one or more classifications on the formatted version of the heterogeneous data to provide the one or more insights. In one or more embodiments, performing the deep learning process includes mapping the heterogeneous data to multidimensional word embeddings using a recurrent neural network. In one or more embodiments, performing the deep learning process includes using a network of gated recurrent units of the recurrent neural network to provide the one or more insights. Performing the deep learning process results in one or more technical improvements, such as, but not limited to, enhancing the functionality of the computing device and / or improving the accuracy of the data provided to the computing device. In one or more embodiments, method 1300 further includes retraining one or more portions of the recurrent neural network based on the one or more insights. Retraining one or more portions of the recurrent neural network results in one or more technical improvements, such as, but not limited to, improving the accuracy of the recurrent neural network. In one or more embodiments, method 1300 further includes determining one or more actions using a scoring model based on different metrics from the iteration history of the deep learning process. Using the scoring model results in one or more technical improvements, such as, but not limited to, extending the capabilities of the computing device and / or improving the accuracy of data provided to the computing device.

[0115] In some exemplary embodiments, some of the operations herein may be modified or further enhanced as described below. Furthermore, in certain embodiments, additional optional operations may be included. It should be appreciated that each of the modifications, optional additions, or enhancements described herein may be included with the operations herein, either alone or in combination with any other of the features described herein.

[0116] FIG. 14 illustrates an exemplary system 1400 that may execute the techniques presented herein. FIG. 14 is a simplified functional block diagram of a computer that may be configured to execute the techniques described herein, according to an exemplary embodiment of the present disclosure. Specifically, the computer (or “platform,” since the computer may not be a single physical computer infrastructure) may include a data communication interface 1460 for packet data communication. The platform may also include a central processing unit (CPU) 1420 in the form of one or more processors for executing program instructions. The platform may include an internal communication bus 1410, and the platform may also include program and / or data storage for various data files processed and / or communicated by the platform, such as ROM 1430 and RAM 1440, although the system 1400 may receive programming and data via network communication. The system 1400 may also include input / output ports 1450 for connecting with input / output devices such as a keyboard, mouse, touchscreen, monitor, display, etc. Of course, various system functions may be implemented in a distributed manner across several similar platforms to distribute processing loads. Alternatively, the system may be implemented by suitable programming of one computer hardware platform.

[0117] 15 illustrates an exemplary user interface 1500 in accordance with one or more embodiments of the present disclosure. In one or more embodiments, the user interface 1500 is an interactive dashboard rendered via a display of a computing device (e.g., computing device 402). The user interface 1500 facilitates data optimization and / or data mapping for the disparate data 314 stored in one or more data sources 316. In one or more embodiments, field mapping 1502 is performed to provide data fluidity for the disparate data 314 stored in the one or more data sources 316. In one example, the disparate data 314 stored in the one or more data sources 316 includes data from five data sources and / or data associated with 1568 auto-populated columns. Further, in one example, the field mapping 1502 is associated with field mapping for 489 columns of data. In one or more embodiments, the user interface 1500 includes an interactive user interface element 1504 that initiates field mapping associated with the data optimization computer system 302 (e.g., initiates generation of the request 320) in accordance with one or more embodiments disclosed herein.

[0118] 16 illustrates an exemplary user interface 1600 according to one or more embodiments of the present disclosure. In one or more embodiments, the user interface 1600 is an interactive dashboard rendered via a display of a computing device (e.g., computing device 402). The user interface 1600 facilitates field mapping for disparate data 314 stored in one or more data sources 316. In one or more embodiments, the one or more data sources 316 include a first data source (e.g., Source Name A) associated with a first source type (e.g., Source Type A), a second data source (e.g., Source Name B) associated with a second source type (e.g., Source Type B), a third data source (e.g., Source Name C) associated with a third source type (e.g., Source Type C), a fourth data source (e.g., Source Name D) associated with the third source type (e.g., Source Type C), and a fifth data source (e.g., Source Name E) associated with the fourth source type (e.g., Source Type D). In one or more embodiments, the field mapping associated with the user interface 1600 is accomplished via a data optimization computer system 302 according to one or more embodiments disclosed herein. In one or more embodiments, the field mapping associated with the user interface 1600 is performed in a reduced amount of time (e.g., seconds, minutes, hours, days, or weeks) compared to conventional data processing systems.

[0119] FIG. 17 illustrates an exemplary user interface 1700 in accordance with one or more embodiments of the present disclosure. In one or more embodiments, the user interface 1700 is an interactive dashboard rendered via a display of a computing device (e.g., computing device 402). The user interface 1700 facilitates field mapping for the heterogeneous data 314 stored in one or more data sources 316. In one or more embodiments, the field mapping associated with the user interface 1700 is implemented via a data optimization computer system 302 in accordance with one or more embodiments disclosed herein. In one or more embodiments, the field mapping associated with the user interface 1700 is performed with respect to source columns and / or target columns of the heterogeneous data 314 stored in one or more data sources 316. In one or more embodiments, the user interface 1700 provides recommendations 1702 for particular source columns (e.g., recommendations for the record_type source column, etc.). In one or more embodiments, the field mapping associated with the user interface 1700 is performed based on a target dictionary associated with a dataset category, logical name, physical name, and / or other information about the target column.

[0120] The foregoing method descriptions and process flow diagrams are provided merely as illustrative examples and do not require or imply that the steps of the various embodiments must be performed in the order presented. As will be understood by one of ordinary skill in the art, the order of steps in the foregoing embodiments may be performed in any order. Words such as "thereafter," "then," and "next" do not limit the order of the steps. These words are merely used to guide the reader through the method descriptions. Furthermore, any reference to claim elements in the singular, for example, using the articles "a," "an," or "the," should not be construed as limiting the element to the singular.

[0121] It should be understood that "one or more" includes functions performed by one element, functions performed by two or more elements, e.g., in a distributed manner, several functions performed by one element, several functions performed by several elements, or any combination of the above.

[0122] Furthermore, while terms such as "first," "second," and the like are sometimes used herein to describe various elements, it should be understood that these elements should not be limited by these terms. These terms are used only to distinguish one element from another. For example, a first contact can be referred to as a second contact, and similarly, a second contact can be referred to as a first contact, without departing from the scope of the various embodiments described. Although a first contact and a second contact are both contacts, they are not the same contact.

[0123] The terminology used in the description of the various embodiments set forth herein is for the purpose of describing particular embodiments only and is not intended to be limiting. When used in the description of the various embodiments set forth and in the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly dictates otherwise. It will also be understood that the term "and / or," as used herein, refers to and encompasses any and all possible combinations of one or more of the associated listed items. It will be further understood that the terms "includes," "including," "comprises," and / or "comprising," when used herein, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0124] As used herein, the term "if" is optionally interpreted to mean "when" or "at the time" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "when determined" or "when [the stated condition or event] is detected" is optionally interpreted to mean "at the time of determining" or "in response to determining" or "on detection of [the stated condition or event]" or "in response to detecting [the stated condition or event]," depending on the context.

[0125] The systems, apparatus, devices, and methods disclosed herein are described in detail by way of examples and with reference to the drawings. The examples discussed herein are merely examples and are provided to aid in the description of the apparatus, devices, systems, and methods described herein. None of the features or components shown in the drawings or discussed below should be construed as essential to any particular implementation of any of these apparatus, devices, systems, or methods, unless specifically designated as essential. For readability and clarity, some components, modules, or methods may only be described in connection with particular figures. In this disclosure, references to specific techniques, configurations, etc. are either related to the specific examples presented or are merely general descriptions of such techniques, configurations, etc. References to specific details or examples are not intended to, and should not be construed as essential or limiting, unless specifically so designated. Failure to specifically describe a combination or subcombination of elements should not be understood as indicating that any combination or subcombination is not possible. It will be understood that modifications to the disclosed and described examples, arrangements, configurations, components, elements, apparatus, devices, systems, methods, etc. can be made and may be desired for particular applications. Also, with respect to any method described, whether the method is described in conjunction with a flow diagram or not, unless otherwise specified or required by context, it will be understood that any explicit or implicit ordering of steps performed in performing the method does not imply that the steps must be performed in the order presented, but instead may be performed in a different order or in parallel.

[0126] Throughout this disclosure, references to components or modules generally refer to items that can be logically grouped together to perform a function or a group of related functions. Like reference numbers are intended to generally refer to the same or similar components. Components and modules can be implemented in software, hardware, or a combination of software and hardware. The term "software" is used expansively to include not only executable code, e.g., machine-executable or machine-interpretable instructions, but also data structures, data stores, and computing instructions stored in any suitable electronic format, including firmware and embedded software. The terms "information" and "data" are used expansively to include a wide variety of electronic information, including executable code, content such as text, video data, and audio data, among other things, as well as various codes or flags. The terms "information," "data," and "content" may be used interchangeably where the context allows.

[0127] The hardware used to implement the various example logic, logic blocks, modules, and circuits described in connection with aspects disclosed herein may include a general-purpose processor, a digital signal processor (DSP), a special-purpose processor such as an application-specific integrated circuit (ASIC) or field-programmable gate array (FPGA), programmable logic devices, discrete-gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general-purpose processor may be a microprocessor, but alternatively, the processor may be any processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, e.g., a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration. Alternatively, or additionally, some steps or methods may be performed by circuitry specific to a given function.

[0128] In one or more exemplary embodiments, the functions described herein may be implemented by dedicated hardware or a combination of hardware programmed by firmware or other software. In implementations relying on firmware or other software, the functions may be performed as a result of execution of one or more instructions stored on one or more non-transitory computer-readable media and / or one or more non-transitory processor-readable media. These instructions may be embodied by one or more processor-executable software modules resident on one or more non-transitory computer-readable or processor-readable storage media. A non-transitory computer-readable or processor-readable storage medium, in this regard, may include any storage medium that can be accessed by a computer or processor. By way of example and not limitation, such non-transitory computer-readable or processor-readable media may include random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, disk storage, magnetic storage devices, etc. As used herein, disk storage devices include compact discs (CDs), laser discs, optical discs, digital versatile discs (DVDs), floppy disks, and Blu-ray discs™, or other storage devices that use lasers to store data magnetically or optically. Combinations of the above types of media are also included within the scope of the terms non-transitory computer-readable medium and processor-readable medium. Additionally, any combination of instructions stored on one or more non-transitory processor-readable medium or computer-readable medium may be referred to herein as a computer program product.

[0129] Many modifications and other embodiments of the inventions described herein will come to mind to one skilled in the art to which these inventions pertain having the benefit of the teachings presented in the foregoing descriptions and the associated drawings. The drawings illustrate only certain components of the devices and systems described herein, it being understood that various other components may be used in conjunction with the supply management system. It is understood, therefore, that the invention is not limited to the specific embodiments disclosed, and that modifications and other embodiments are intended to be included within the scope of the appended claims. Furthermore, the steps in the methods described above need not necessarily occur in the order depicted in the accompanying drawings; in some cases, one or more of the depicted steps may occur substantially simultaneously, or additional steps may be included. Although specific terms are employed herein, they are used in a generic and descriptive sense only and not for purposes of limitation.

[0130] It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

Claims

1. 1. A system comprising: one or more processors; Memory and one or more programs stored in the memory, the one or more programs including instructions, the instructions receiving data from a plurality of data sources, the data relating to at least one of an asset or a service of the enterprise; formatting the data using one or more predetermined formats; receiving a request to obtain one or more insights about the formatted data, the formatted data being organized as an ontology tree structure that captures relationships between two or more portions of the data as a hierarchy of nodes and connections, and each of the one or more insights being associated with the formatted data, the request comprising: an insight descriptor describing a goal of the one or more insights, the goal being to provide optimization for at least one of the assets or the services of the enterprise; In response to said request, correlating aspects of the formatted data with one or more machine learning models to provide the one or more insights, the one or more insights comprising one or more actionable recommendations for sustained peak performance of the enterprise, the correlated aspects being determined by the goals and relationships between the aspects of the formatted data, the relationships being captured by the ontology tree structure; performing one or more actions based on the one or more insights; displaying the one or more actions in a user-interactive electronic interface to render a visual representation of the one or more insights; A system that is configured to:

2. The one or more programs further include instructions, the instructions comprising: obtaining the data received from the plurality of data sources; formatting one or more portions of the data based on the ontology tree structure, said formatting providing the formatted data associated with a defined format; and determining one or more mapping recommendations for the formatted data, the one or more mapping recommendations including predicted column name data fields corresponding to the one or more portions of the data.

3. The one or more programs further include instructions, the instructions comprising: determining one or more data fields in the data received from the plurality of data sources, the one or more data fields comprising at least one of a data element, a word, or a phrase associated with the data; determining one or more incomplete data fields within the data received from the plurality of data sources; determining whether at least one data field of the one or more incomplete data fields of a first data source is identical to at least one data field of a second data source, wherein data corresponding to the at least one data field of the second data source is complete; and adding the data from the at least one data field of the second data source to the at least one data field of the one or more incomplete data fields in accordance with the determination.

Images