Service resource dynamic optimization configuration method and system based on deep learning prediction

Abstract

The invention belongs to the technical field of service resource allocation, and provides a service resource dynamic optimization configuration method and system based on deep learning prediction, and the method comprises the steps: obtaining multi-source data, and carrying out the standardization processing of the multi-source data through a predefined data template; key associated fields of the standardized multi-source data are extracted, and a data graph is constructed according to business logic; extracting service peak data associated with resource allocation in the data graph, and performing feature extraction on the service peak data; utilizing a pre-trained deep learning model to process the extracted time sequence features, the associated features and the burst features, and predicting business data of a target time period; and according to the predicted service data, determining an allocation scheme of each type of resources according to a mapping rule base of a preset service scene, a resource type and a quantitative standard relationship. According to the invention, accurate prediction of services can be realized, and accurate optimal configuration of service resources is further realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of business resource allocation, and specifically relates to a method and system for dynamically optimizing business resource configuration based on deep learning prediction. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Enterprise Resource Planning (ERP) software, as the core carrier of an enterprise's business processes, accumulates massive amounts of business data (such as order data, production data, inventory data, financial data, etc.). The generation and flow of this data directly reflects the characteristics of the enterprise's business load (such as surges in order volume, centralized approval of production work orders, peaks in logistics documents, etc.). However, existing ERP systems have significant limitations in data value mining and decision support.

[0004] First, the ERP system adopts a modular architecture, and each module runs independently, resulting in decentralized data storage and incompatibility with the data formats and protocols of external systems, forming "data islands" and making it difficult to form a global business view. As a result, it is impossible to accurately capture the related factors of business peaks. For example, sales order data cannot be linked with production equipment capacity data and raw material inventory data in real time, resulting in a lack of global data support for production planning.

[0005] On the other hand, the ERP system's predictions rely on manual experience or simple statistical methods (such as historical averages), and on the historical business volume of a single module. It does not consider the correlation between multi-source data and cannot adapt to the impact of sudden factors in industrial scenarios. The prediction error is large, resulting in low prediction accuracy. Due to inaccurate predictions, enterprises often face the problem of insufficient resources during business peaks and idle resources during business troughs, resulting in low operational efficiency and waste of resource costs. Summary of the Invention

[0006] In order to solve the above problems, the present invention proposes a method and system for dynamic optimization configuration of business resources based on deep learning prediction. The present invention is based on the deep fusion of multi-source data, takes deep learning prediction as the core, and aims at closed-loop resource allocation, constructs a full-link intelligent mechanism, and can realize accurate prediction of business, and then realize accurate optimization configuration of business resources.

[0007] According to some embodiments, the present invention adopts the following technical solutions: A method for dynamically optimizing business resource configuration based on deep learning prediction includes the following steps: Acquire multi-source data, and perform standardization processing on the multi-source data using a predefined data template; Extract key related fields from standardized multi-source data and build data graphs based on business logic; Extract business peak data associated with resource allocation from the data map and perform feature extraction on the business peak data. The extracted features include time series features, correlation features, and burst features. Use pre-trained deep learning models to process extracted time series features, correlation features, and burst features to predict business data for the target period. Based on the predicted business data, the allocation plan for various types of resources is determined according to the preset mapping rule library of business scenarios, resource types and quantitative standard relationships.

[0008] As an optional implementation method, the process of acquiring multi-source data includes: predefining an interface template library, utilizing the interface template library to connect to an external system, and acquiring external related data; Leverage the low-code configuration platform to accept registration of new data sources through drag-and-drop operations.

[0009] Furthermore, the external related data includes equipment operation data, warehouse operation efficiency data, customer demand and credit rating, equipment real-time status data, supply cycle and quality data and market environment data.

[0010] As an optional implementation method, the process of extracting key associated fields of standardized multi-source data and constructing a data graph according to business logic includes: extracting key associated fields, which include order number, material code and timestamp, and constructing a data graph according to business processes and business logic, wherein the data graph includes document characteristics, business scenarios and external influencing factors.

[0011] As an optional implementation, the process of extracting business peak data associated with resource allocation in the data map includes: extracting sales order peak data, raw material demand peak data, production load peak data, and inventory turnover peak data.

[0012] As an optional implementation, the process of extracting features from business peak data includes: extracting time series features, including periodic peak patterns of historical business data and peak evolution in long-term trends; Extract relevant features, including cross-business correlations that drive peaks and the positive correlation between equipment failures and production work order backlogs; Extract burst features, including key factors that trigger aperiodic peaks.

[0013] As an optional implementation, the process of processing the extracted time series features, correlation features, and burst features using a pre-trained deep learning model includes: The pre-trained deep learning model includes a convolutional neural network and a long short-term memory network, wherein the convolutional neural network is used to extract local features, and the long short-term memory network is used to deeply capture the temporal dependencies of business peak data, accurately identify the periodic patterns of historical peaks, and introduce an attention mechanism to enhance features with significant peak impacts.

[0014] As a further implementation method, the pre-trained deep learning model also introduces enterprise resource planning business rules to correct the output of the pre-trained deep learning model so that the output complies with the enterprise resource planning business rules.

[0015] As a further implementation method, the parameters of the pre-trained deep learning model are dynamically adjusted based on the latest business data of the enterprise resource planning after a set time period.

[0016] As an optional implementation method, the process of determining the allocation plan for various types of resources based on the predicted business data and in accordance with the preset mapping rule library of business scenarios, resource types and quantitative standard relationships includes: the preset mapping rule library of business scenarios, resource types and quantitative standard relationships contains resources associated with various business peaks, as well as strategies for matching calculations of various types of resources. Based on the predicted business data and in accordance with the mapping rule library, the quantity of various types of resources that need to be matched is calculated to form an allocation plan, and it is determined whether the quantity of various types of resources in the allocation plan exceeds the set threshold. If not, it is automatically sent to the corresponding business module. If so, a high-risk decision work order is generated for approval.

[0017] A business resource dynamic optimization configuration system based on deep learning prediction, comprising: a multi-source data processing module configured to acquire multi-source data and perform standardization processing on the multi-source data using a predefined data template; The data graph construction module is configured to extract key related fields of standardized multi-source data and construct a data graph according to business logic; A peak feature extraction module is configured to extract business peak data associated with resource allocation from the data map and perform feature extraction on the business peak data. The extracted features include time series features, correlation features, and burst features. The business prediction module is configured to use a pre-trained deep learning model to process the extracted time series features, correlation features, and burst features to predict business data for the target period; The resource allocation module is configured to determine the allocation plan for various types of resources based on the predicted business data and the mapping rule library of preset business scenarios, resource types and quantitative standard relationships.

[0018] Compared with the prior art, the present invention has the following beneficial effects: The present invention adopts a combination of predefined interface templates and low-code configuration to achieve direct access between ERP internal data and external data, solving the problems of scattered data sources and heterogeneous formats; and constructs association rules based on the ERP core business process, and automatically establishes an association map of cross-system data through key fields such as order number and material code, forming a fusion data pool that includes document characteristics, business scenarios, and external influences, ensuring deep binding of data and business scenarios; through format conversion, it ensures the consistency and accuracy of the fusion data, providing high-quality data support for subsequent links. Based on fused data, the present invention constructs a prediction model that has both time series analysis capabilities and business adaptability, accurately captures the laws of business peaks, and mines three types of key features from multi-source data: time series features, correlation features, and burst features, comprehensively covering peak driving factors; and when making predictions, it adopts a model that combines convolutional neural networks, long- and short-term neural networks, attention mechanisms, and business rule corrections. It utilizes the local feature extraction capabilities of convolutional neural networks and the capture of time series dependencies through long- and short-term neural networks, the attention mechanism strengthens the weights of key features, and the business rules are calibrated to adapt to the special laws of industrial scenarios, significantly improving prediction accuracy.

[0019] The present invention dynamically adjusts model parameters based on the latest business data at set time intervals, quickly responds to emergencies such as urgent orders and equipment failures, controls the prediction error of core indicators within the set value, and provides accurate time and magnitude basis for resource allocation.

[0020] The present invention converts prediction results into executable resource allocation instructions to form a dynamic optimization closed loop of prediction-execution-feedback, and establishes a rule library of business scenarios-resource types-quantification standards based on the predicted business data to achieve accurate matching of resource requirements and business pressure.

[0021] The present invention lays the foundation through data fusion, realizes the progressive process of value through deep learning accurate prediction and closed-loop allocation, realizes the upgrade of enterprise resource planning from passive recording to active decision-making, and improves the efficiency of enterprise resource utilization and the level of operational intelligence.

[0022] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0024] Figure 1 is a flow chart of an embodiment; Figure 2This is a schematic diagram of the deep learning model structure of an embodiment. DETAILED DESCRIPTION

[0025] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0026] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0027] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0028] In the absence of conflict, the embodiments and features in the embodiments of this application can be combined with each other.

[0029] Example 1 As mentioned in the background technology, the modular architecture of existing ERP systems leads to decentralized data storage and is incompatible with external system data formats and protocols, which easily leads to data silos.

[0030] Moreover, the data integration of existing ERP systems relies on customized development (such as hard-coded interfaces). When new data sources are added (such as introducing new supplier systems or adding new production sensors) or the data source format changes, the interface needs to be redeveloped, which takes weeks or even months. This makes it impossible to adapt to dynamic needs such as production line adjustments and supply chain changes in industrial enterprises.

[0031] Existing ERP systems often rely on manual experience or simple statistical methods (such as historical averages) for business forecasting, and on historical business volume within a single module. They fail to consider the correlations between multiple data sources (such as the impact of raw material price fluctuations on procurement volume or the impact of equipment failure rates on production schedules). Furthermore, these systems are unable to adapt to the impact of unexpected factors in industrial scenarios (such as urgent orders and equipment failures). This leads to large forecast errors and low forecast accuracy (for example, forecast deviations exceeding 30% during peak order volumes). Due to inaccurate forecasts, companies often face resource shortages during peak business periods (such as shortages of approval personnel and logistics capacity) and idle resources during low business periods, resulting in low operational efficiency (such as delayed order delivery) and wasted resource costs. Furthermore, the decision-making process is disconnected from the forecast results. For example, when a sales order peak is predicted, production plan adjustments and logistics capacity allocation cannot be automatically triggered, leading to resource shortages or idle resources.

[0032] Existing ERP systems rely heavily on built-in models from product design for data processing. As businesses grow, these models' limitations become increasingly apparent. Furthermore, due to the ERP system's inherent size and historical burden, data models and processing capabilities are difficult to upgrade quickly, making them unable to adapt to modern enterprise management systems and models.

[0033] In order to solve the above problems such as data silos, prediction lag, and inefficient resource allocation, this embodiment provides a method for dynamic optimization of business resources based on deep learning prediction, such as Figure 1 As shown, the following steps are included: Acquire multi-source data, and perform standardization processing on the multi-source data using a predefined data template; Extract key related fields from standardized multi-source data and build data graphs based on business logic; Extract business peak data associated with resource allocation from the data map and perform feature extraction on the business peak data. The extracted features include time series features, correlation features, and burst features. Use pre-trained deep learning models to process extracted time series features, correlation features, and burst features to predict business data for the target period. Based on the predicted business data, the allocation plan for various types of resources is determined according to the preset mapping rule library of business scenarios, resource types and quantitative standard relationships.

[0034] In this embodiment, the process of acquiring multi-source data includes: taking full coverage, flexible adaptation, and deep association as core design principles, adopting a multi-dimensional, full-scenario access strategy at the data collection level, and comprehensively integrating two types of key data resources.

[0035] First, the core business data within the ERP system covers all modules, including procurement management (order quantity, delivery cycle, supplier information), production execution (work order process route, capacity requirements, production progress), sales management (customer information, order demand, delivery deadline), inventory control (material coding, inventory level, warehouse location distribution) and financial management (cost accounting, budget execution, accounts receivable and payable), etc., and fully covers the basic attributes and core business fields such as business document number, timestamp, status, etc.

[0036] Second, external related data includes equipment operation data of the manufacturing execution system (such as equipment overall efficiency, failure rate, production load, etc.), warehouse operation efficiency of the warehouse management system (such as turnover rate, warehousing and outbound time, etc.), customer needs and credit ratings of customer relationship management, and real-time status of equipment collected by IoT sensors (such as temperature, energy consumption, operating parameters, etc.), supply cycle and quality data of the supplier system, and even external market environment data (such as industry promotion nodes, raw material price fluctuations, changes in industry requirements, etc.).

[0037] In addition, in this embodiment, efficient docking is achieved through a three-layer adaptation mechanism: The first is a predefined interface template library that develops standardized interfaces for mainstream external systems and supports one-click access configuration. The second is a dynamic protocol adaptation engine, where users select the corresponding data transmission protocol through a visual interface, and the corresponding format parser and time series data converter are automatically loaded without manual coding; The third is a low-code configuration platform, where business personnel can complete the registration of new data sources (such as filling in access address, authentication information, update frequency and other parameters) by dragging and dropping, compressing the traditional access cycle that takes several weeks to within 1 hour, and flexibly responding to the dynamic changes in data sources in scenarios such as industrial enterprise production line adjustments and supplier changes.

[0038] Then, the data is fused and standardized.

[0039] Through the dynamic format conversion mechanism, data from different sources is standardized based on preset templates and custom rules (such as unit conversion rules and unified date format), ensuring consistency in field naming, data type, and timestamp format.

[0040] In this embodiment, it also relies on key associated fields such as order number, material code, timestamp, etc., and automatically constructs an associated graph of multi-source data according to business logic such as "sales order → production work order → raw material procurement → supplier", "production work order → equipment resources → energy consumption data", forming a fusion data pool that includes "document characteristics - business scenarios - external influencing factors", forming a visual business data graph, and supporting the tracing of upstream and downstream data of any business node.

[0041] Through the above solution, this embodiment can achieve "plug and play" access to multi-source data inside and outside the ERP, solve the problems of scattered data sources and heterogeneous formats, and realize deep integration from isolated data to full-link business data, providing complete, consistent, and relevant high-quality data support for subsequent dynamic prediction and intelligent decision-making.

[0042] During the data prediction process, pre-trained deep learning models are used to accurately predict business peak data or determine the peak of business data.

[0043] The deep learning prediction layer uses multi-source fusion data as its core input. Through targeted deep learning models and dynamic optimization mechanisms, it can accurately predict the peak of core business data in the ERP system, providing a decision-making basis for enterprises to respond to business fluctuations in advance.

[0044] Extract core business data from the ERP system that is closely related to resource allocation, including sales order peaks (e.g., peak order volume dates, volume levels, and the proportion of orders from key customers over the next 7 / 30 days), raw material demand peaks (e.g., peak procurement volume and potential shortfalls within 15 days), production load peaks (e.g., peak work order saturation rates for each production line within a week, critical equipment utilization points), and inventory turnover peaks (e.g., peak consumption periods for key materials and out-of-stock warning windows). This data directly reflects where business pressures are concentrated, and its predictions serve as the core basis for companies to proactively allocate resources.

[0045] Targeted mining of three types of features that are strongly correlated with business peaks is conducted from the fused data: First, time series features: the periodic peak patterns of historical business data (such as a surge in sales orders every Monday and a concentrated outbreak of production work orders at the end of a quarter), and the evolution of peaks in long-term trends (such as the annual increase in peak season order peaks); second, correlation features: the cross-business correlation that drives peaks (such as a peak in purchase orders caused by plummeting raw material prices and a chain peak in sales and logistics orders caused by promotional activities), and the positive correlation between equipment failures and production work order backlogs; and third, sudden features: the key factors that trigger non-periodic peaks (such as a sharp increase in the proportion of emergency orders, a concentrated replenishment of production work orders due to delayed delivery by suppliers, and temporary order increases from customers).

[0046] In addition, this embodiment also uses a deep learning model to optimize the accuracy of peak prediction.

[0047] This embodiment adopts a model architecture of convolutional neural network + attention mechanism + long short-term memory network + business rule calibration.

[0048] Specifically, such as Figure 2 As shown, it includes the input layer, convolutional neural network layer, long short-term memory network layer, attention layer, fully connected layer and output layer arranged in sequence.

[0049] The input layer receives the normalized and standardized multi-source data, which can be divided into the data set {X1, X2, ..., X n}.

[0050] The convolutional neural network layer leverages the local feature extraction capabilities of the convolutional neural network (CNN), and the long short-term memory network layer uses several long short-term memory (LSTM) networks to deeply capture the temporal dependencies of business data and accurately identify cyclical patterns of historical peaks (such as the annual sales peak).

[0051] The convolutional neural network includes a convolution layer and a pooling layer, which can be implemented using existing technologies and will not be described in detail here. The long short-term memory network can also be implemented using existing technologies and will not be described in detail here.

[0052] The attention layer introduces an attention mechanism to focus on features that have a significant impact on peaks (such as the weight of promotional activities on sales order peaks, the impact of equipment maintenance plans on production load peaks, etc.). This layer can use the Softmax function as the activation function.

[0053] Next, after dense connection through the fully connected layer, it is output through the output layer.

[0054] A calibration layer can also be introduced to incorporate ERP business rules (such as "order volume increased by 30% month-on-month in the two weeks before the Spring Festival" and "peaks formed by concentrated orders from major customers on the 5th of each month") to correct the output of the deep learning model to adapt to regular sudden peaks in industrial scenarios.

[0055] To ensure adaptability to dynamic business, the deep learning model establishes a real-time update mechanism: this embodiment dynamically adjusts the optimization model parameters (such as the weight of the attention mechanism and the model's cycle factor) every 6 hours based on the latest ERP business data (such as newly generated emergency orders, equipment failure records, material inventory information, etc.) to quickly respond to unexpected peaks caused by sudden factors (such as emergency orders and equipment downtime).

[0056] Through this mechanism, the prediction error of core business data peaks can be controlled within 10%, among which the prediction accuracy of key indicators such as sales order peaks and production load critical points is particularly outstanding, providing companies with accurate time and magnitude basis for adjusting production plans, stockpiling raw materials, allocating logistics capacity and other response measures in advance.

[0057] Based on the business peak data output by the deep learning prediction layer, through regular matching, phased strategies and closed-loop feedback mechanisms, dynamic adaptation of resources and business needs is achieved, thereby improving the resource utilization efficiency of the ERP system.

[0058] The resource demand matching mechanism focuses on the precise correspondence between business peaks and resource types: a mapping rule library of business scenarios, resource types, and quantitative standards is established based on prediction results.

[0059] For example, in the mapping rule library, a sales order peak (>500 orders / day) corresponds to logistics capacity and approval personnel resources. Based on the standard configuration of 2 approvers and 3 delivery vehicles for 100 orders, the system requires 2 approvers and 3 delivery vehicles.

[0060] In the mapping rule base, production load peaks (production line saturation rate > 90%) are associated with two resources: equipment and technicians, and are allocated at a ratio of three technicians per device.

[0061] When the raw material demand gap exceeds the set value (for example, the gap quantity is greater than 500 pieces), both procurement and inventory resource responses are triggered, and a purchase order is generated according to the calculation formula of gap quantity × 1.2 safety factor.

[0062] The mapping rule library supports customization according to ERP business processes, such as inventory allocation rules for cross-regional sales (for example, when orders increase by 50%, 30% of safety stock will be allocated from sufficient warehouses).

[0063] The generated dynamic allocation strategies can be executed in layers based on timelines, ensuring timely and proactive resource responses. For example, short-term (1-3 days) strategies, targeting emergency peaks, directly generate ERP executable instructions: These can push approval work orders to idle personnel via the Application Programming Interface (API), mobilize outsourced logistics capacity to fill gaps, and temporarily adjust production line loads (dispatching overwhelmed work orders to idle equipment).

[0064] Medium-term strategies (e.g., 1-4 weeks) focus on planned adjustments: scheduling equipment maintenance in advance based on production peak forecasts (e.g., avoiding breakdowns during peak hours), optimizing inventory layout based on sales trends (e.g., pre-positioning popular products to regional warehouses), and developing technician scheduling plans (e.g., adding 20% ​​staff on peak days).

[0065] Allocation strategy execution adopts two modes: automatic and approval. General standardized operations (such as routine procurement and inventory allocation) are directly implemented by the system, that is, automatic mode, and automatically sent to the corresponding business module for execution. High-risk decisions (such as emergency procurement exceeding the budget and resource allocation exceeding the set value) can be pushed to the ERP workbench for approval and execution.

[0066] In some embodiments, the deployed business data (such as order delivery timeliness rate, equipment utilization rate, purchase arrival rate, etc.) can also be collected in real time and compared with the expected goals.

[0067] For example, if the actual verification of 100 sales orders requires three approvers, the mapping rule base will be automatically updated. For example, if it is found that 500 orders per day require four delivery vehicles, the logistics capacity standard will be adjusted. If the raw material shortage forecast deviates from the actual situation by more than 15%, the safety factor and the parameters of the deep learning model used for prediction will be adjusted in the opposite direction.

[0068] Through this mechanism, the matching degree between resource allocation and business needs is continuously improved, which not only avoids resource shortages during peak hours, but also reduces idle waste during low hours, thus realizing the intelligent and dynamic resource scheduling of the ERP system.

[0069] Example 2 A business resource dynamic optimization configuration system based on deep learning prediction, comprising: a multi-source data processing module configured to acquire multi-source data and perform standardization processing on the multi-source data using a predefined data template; The data graph construction module is configured to extract key related fields of standardized multi-source data and construct a data graph according to business logic; A peak feature extraction module is configured to extract business peak data associated with resource allocation from the data map and perform feature extraction on the business peak data. The extracted features include time series features, correlation features, and burst features. The business prediction module is configured to use a pre-trained deep learning model to process the extracted time series features, correlation features, and burst features to predict business data for the target period; The resource allocation module is configured to determine the allocation plan for various types of resources based on the predicted business data and the mapping rule library of preset business scenarios, resource types and quantitative standard relationships.

[0070] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of one or more computer-usable storage media (including but not limited to disk storage, CD - ROM , optical storage, etc.).

[0071] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1A device that provides the functions specified in a block or multiple blocks.

[0072] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0073] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0074] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made by those skilled in the art that fall within the spirit and principles of the present invention and do not require creative effort are intended to be within the scope of protection of the present invention.

Claims

1. A method for dynamic optimization and configuration of business resources based on deep learning prediction, characterized by: The following steps are involved: Acquire multi-source data, and perform standardization processing on the multi-source data using a predefined data template; Extract key related fields from standardized multi-source data and build data graphs based on business logic; Extract business peak data associated with resource allocation from the data map and perform feature extraction on the business peak data. The extracted features include time series features, correlation features, and burst features. Use pre-trained deep learning models to process extracted time series features, correlation features, and burst features to predict business data for the target period. Based on the predicted business data, the allocation plan for various types of resources is determined according to the preset mapping rule library of business scenarios, resource types and quantitative standard relationships.

2. The method for dynamic optimization and configuration of business resources based on deep learning prediction according to claim 1, characterized in that: The process of acquiring multi-source data includes: pre-defining an interface template library, using the interface template library to connect to external systems, and acquiring external related data; Leverage the low-code configuration platform to accept registration of new data sources through drag-and-drop operations.

3. The method for dynamic optimization and configuration of business resources based on deep learning prediction according to claim 2, characterized in that: The external related data includes equipment operation data, warehouse operation efficiency data, customer demand and credit rating, equipment real-time status data, supply cycle and quality data and market environment data.

4. The method for dynamic optimization and configuration of business resources based on deep learning prediction according to claim 1, characterized in that: The process of extracting key associated fields from standardized multi-source data and constructing a data graph according to business logic includes: extracting key associated fields, which include order numbers, material codes, and timestamps; and constructing a data graph according to business processes and business logic, which includes document features, business scenarios, and external influencing factors.

5. The method for dynamic optimization and configuration of business resources based on deep learning prediction according to claim 1, characterized in that: The process of extracting business peak data associated with resource allocation in the data map includes: extracting sales order peak data, raw material demand peak data, production load peak data, and inventory turnover peak data.

6. The method for dynamic optimization and configuration of business resources based on deep learning prediction according to claim 1, characterized in that: The process of feature extraction for business peak data includes: extracting time series features, including the periodic peak patterns of historical business data and the evolution of peaks in long-term trends; Extract relevant features, including cross-business correlations that drive peaks and the positive correlation between equipment failures and production work order backlogs; Extract burst features, including key factors that trigger aperiodic peaks.

7. The method for dynamic optimization and configuration of business resources based on deep learning prediction according to claim 1, characterized in that: The process of processing the extracted time series features, correlation features, and burst features using the pre-trained deep learning model includes: The pre-trained deep learning model includes a convolutional neural network and a long short-term memory network, wherein the convolutional neural network is used to extract local features, and the long short-term memory network is used to deeply capture the temporal dependencies of business peak data, accurately identify the periodic patterns of historical peaks, and introduce an attention mechanism to enhance features with significant peak impacts.

8. The method for dynamic optimization and configuration of business resources based on deep learning prediction according to claim 7, characterized in that: The pre-trained deep learning model also introduces enterprise resource planning business rules to correct the output of the pre-trained deep learning model so that the output complies with the enterprise resource planning business rules; The parameters of the pre-trained deep learning model are dynamically adjusted based on the latest business data of the enterprise resource planning after a set time period.

9. The method for dynamic optimization and configuration of business resources based on deep learning prediction according to claim 1, characterized in that: The process of determining the allocation plan for various types of resources based on the predicted business data and in accordance with the preset mapping rule library of business scenarios, resource types and quantitative standard relationships includes: the preset mapping rule library of business scenarios, resource types and quantitative standard relationships contains resources associated with various business peaks, as well as strategies for calculating the matching of various types of resources. Based on the predicted business data and in accordance with the mapping rule library, the number of various types of resources that need to be matched is calculated to form an allocation plan. It is determined whether the number of various types of resources in the allocation plan exceeds the set threshold. If not, it is automatically sent to the corresponding business module. If so, a high-risk decision work order is generated for approval.

10. A business resource dynamic optimization configuration system based on deep learning prediction, characterized by: include: a multi-source data processing module configured to acquire multi-source data and perform standardization processing on the multi-source data using a predefined data template; The data graph construction module is configured to extract key related fields of standardized multi-source data and construct a data graph according to business logic; A peak feature extraction module is configured to extract business peak data associated with resource allocation from the data map and perform feature extraction on the business peak data. The extracted features include time series features, correlation features, and burst features. The business prediction module is configured to use a pre-trained deep learning model to process the extracted time series features, correlation features, and burst features to predict business data for the target period; The resource allocation module is configured to determine the allocation plan for various types of resources based on the predicted business data and the mapping rule library of preset business scenarios, resource types and quantitative standard relationships.

Images