A process approval method and system based on active cleaning and fusion of heterogeneous data

Abstract

The present invention proposes a process approval method and system based on active cleaning and fusion of heterogeneous data, including a heterogeneous data cleaning and fusion module, a process panoramic visualization module, a message push module and an interactive decision support module; extracting fusion data by full extraction of ETL tools, incremental extraction of system API interfaces and real-time incremental extraction of users, the system customizes the cleaning entry and actively cleans the data; constructing an approval flowchart according to the approval process data and using a polling mechanism or WebSocket to update the approval process data in real time; configuring interactive message push rules, and integrating a machine learning model through an intelligent early warning engine to predict the delay risk in the approval process and trigger an early warning signal; adjusting the scheduling strategy according to the early warning signal, and sending an OA email message reminder when the early warning mechanism is triggered. The present invention reduces the dependence on different business system process engines, can quickly adjust the data processing approval process according to business needs, and better perform business decision-making and process management.

Description

Technical Field

[0001] The present invention relates to the technical field of data management and data processing, and in particular to a process approval method and system based on active cleaning and fusion of heterogeneous data. Background Art

[0002] At present, the traditional BPM model (Business Process Management) usually requires that the processes of different business systems be integrated into a unified engine. However, when the business systems of a group enterprise come from different suppliers, it may involve solving supplier-specific compatibility issues, integration difficulties and additional customized development costs, which limits the flexibility and diversity of the enterprise in the field of informatization. Therefore, the traditional BPM model (Business Process Management) faces integration challenges in the unified process engine in a multi-supplier environment.

[0003] In addition, it lacks flexibility and scalability in the face of rapidly changing business needs, making it difficult to integrate processes to adapt to the needs of new business scenarios. At the same time, the complexity and degree of customization of business systems make it difficult to quickly and smoothly integrate process engines, limiting the unified management of process data and the consistency of processes. The implementation and maintenance of the BPM model (Business Process Management) usually requires professional skills and a lot of customization work, which is not only complex but also leads to high implementation and maintenance costs.

[0004] In addition, the data of various business systems cannot be easily integrated, and the consistency and accuracy of data between various business systems cannot be guaranteed. It relies on a completely unified process engine and cannot quickly adjust the data processing and approval process according to business needs.

[0005] In view of the above problems, the present invention proposes a process approval method and system based on active cleaning and fusion of heterogeneous data. Summary of the invention

[0006] The present invention proposes the following technical solutions to address one or more technical deficiencies in the above-mentioned prior art.

[0007] Based on the first aspect of the present invention, a process approval system based on active cleaning and fusion of heterogeneous data is proposed, including a heterogeneous data cleaning and fusion module, a process panoramic visualization module, a message push module and an interactive decision support module; wherein,

[0008] The heterogeneous data cleaning and fusion module is configured to extract and fuse data by full extraction using ETL tools, incremental extraction using the system API interface, and real-time incremental extraction using users. The system customizes the cleaning entry and actively cleans the data according to the time period or unique identifier.

[0009] The specific steps of extracting data for fusion, defining a system-defined cleaning entry, and actively cleaning data according to a time period or a unique identifier are as follows:

[0010] Use ETL tools to fully extract the initialized approval process data. If the approval process data is not initialized, obtain heterogeneous data from different business systems and register them in the heterogeneous system pool. After data source adaptation, extract the heterogeneous data in batches according to conditions and put them into the data extraction pool.

[0011] If there is data in the data extraction pool, the heterogeneous data or the approval process data is format-standardized, and the null values, abnormal values ​​and duplicate values ​​are cleaned and put into a unified data model conversion center;

[0012] The model conversion center maps fields of different data extraction methods through a mapping table in a database, configures an API service-type extraction method, and writes a program template to template the script tool program;

[0013] Through the model conversion center, the heterogeneous data and the approval process data are matched with each other in fields, unified in data coding, and updated;

[0014] If the approval process is not completed, the heterogeneous data and the approval process data are placed in the unfinished data pool. If the automatic cleaning task time arrives, the system automatically extracts, cleans and merges. Otherwise, the user actively extracts, cleans and merges and extracts data according to data identifiers and places them in the data extraction pool.

[0015] The process panorama visualization module is configured to: construct an approval process diagram according to the approval process data and update the approval process data in real time using a polling mechanism or WebSocket;

[0016] The approval process data includes approval node, approval status, timestamp and approver information;

[0017] The message push module is configured to: configure interactive message push rules, integrate machine learning models through an intelligent early warning engine, and predict delay risks in the approval process and trigger early warning signals;

[0018] The interactive decision support module is configured as follows: the system adjusts the scheduling strategy according to the early warning signal, and sends an OA email message reminder when the early warning mechanism is triggered.

[0019] Extracting data using the system API interface incremental extraction method not only improves data extraction efficiency, but also reduces the load on heterogeneous systems.

[0020] The intelligent early warning engine can predict potential delays in the approval process, identify bottlenecks in the approval process, and provide intervention suggestions, allowing users to take measures in advance and reduce the risk of delays.

[0021] Furthermore, the message push rules include approval node timeout threshold, push conditions and account dynamic management;

[0022] The fields of the mapping table include a business system identifier, an extraction method, an original data table field name, and a new data table field name.

[0023] Based on the second aspect of the present invention, a method for process approval using a process approval system based on active cleaning and fusion of heterogeneous data is also proposed, and the method comprises the following steps:

[0024] The approval process data is input into the system. The heterogeneous data cleaning and fusion module uses the interface opened by the business system to extract the approval process data from different business systems in the form of API services, and then configures the corresponding scheduled tasks through the ETL tool, and extracts the approval process data for cleaning and fusion. The process panoramic visualization module constructs an approval process diagram based on the approval process data and displays the approval process nodes and the status of each link;

[0025] The system determines whether the current approval process is stagnant. If the current approval process is not stagnant, the system uses the machine learning model to predict whether the current approval process will be delayed. If the current approval process is stagnant or the machine learning model predicts that the current approval process will be delayed, the system triggers an early warning mechanism and sends an OA email message reminder.

[0026] If the machine learning model predicts that the approval will not be delayed, the priority of the current approval process is calculated based on the load of the current approvers and the approval process data sorted by priority is generated, and then the approval is performed according to the priority of the approval process data.

[0027] Using ETL tools, you can integrate, clean and convert data from different business systems, integrate and manage data in a unified manner, make the approval process more flexible and adaptable, and choose the most appropriate process engine based on your own business to optimize the approval process.

[0028] Furthermore, the approval process data is input into the machine learning model and the historical approval process data is used to train the machine learning model. Specifically, the format of the approval process data is converted through data preprocessing and feature engineering, and the parameters of the machine learning model are optimized using cross-validation technology and grid search technology;

[0029] Among them, the machine learning model includes an LSTM long short-term memory network model and an XGBoost regression tree enhancement model.

[0030] Furthermore, the LSTM long short-term memory network model monitors the validation set during training, and if the loss of the validation set does not decrease as the training progresses, or the training set is overfitted, the parameters of the LSTM long short-term memory network model are modified to optimize the model configuration;

[0031] Modifying the parameters of the LSTM long short-term memory network model includes reducing the learning rate, and increasing the batch size, hidden layer neurons, number of iterations, and dropout rate.

[0032] Furthermore, the XGBoost regression tree boosting model improves the verification accuracy by reducing the learning rate, increasing the maximum depth of the regression tree, and increasing the minimum number of samples and regularization parameters for regression tree construction when the XGBoost regression tree boosting model is overfitting; and prevents further overfitting by reducing the column sampling ratio and the row sampling ratio; and

[0033] When the training time of the XGBoost regression tree enhancement model is too long, the regression tree is prevented from splitting by increasing the minimum number of samples.

[0034] Based on the third aspect of the present invention, a computer storable medium is further proposed, on which one or more computer programs are stored, and when the one or more computer programs are executed by a computer processor, the above method is implemented.

[0035] The technical effect of the present invention is that the present invention effectively integrates and uniformly manages approval process data, uses data pipelines and ETL tools to build a more flexible and adaptable approval process system, and comprehensively cleans and integrates data in real time. It does not require a completely unified process engine, reduces dependence on process engines of different business systems, and can quickly adjust data processing approval processes according to business needs, making data flow smoother, integrating different business systems and data sources, and better making business decisions and process management. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Other features, objects and advantages of the present application will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings.

[0037] Figure 1 It is a framework diagram of a process approval system based on active cleaning and fusion of heterogeneous data provided according to an embodiment of the present invention.

[0038] Figure 2 It is an overall flow chart of a process approval method based on active cleaning and fusion of heterogeneous data provided according to an embodiment of the present invention.

[0039] Figure 3 It is a specific flow chart for extracting approval process data for cleaning and integration according to an embodiment of the present invention.

[0040] Figure 4 It is a schematic diagram of the structure of a computer system of an electronic device suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0042] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0043] Figure 1 The invention shows a process approval system based on active cleaning and fusion of heterogeneous data, including a heterogeneous data cleaning and fusion module a, a process panoramic visualization module b, a message push module c and an interactive decision support module d.

[0044] In a specific embodiment, the heterogeneous data cleaning and fusion module a is configured to: extract and fuse data by full extraction using ETL tools, incremental extraction using system API interfaces, and real-time incremental extraction using users; the system customizes cleaning entry and actively cleans data based on time periods or unique identifiers;

[0045] The specific steps of extracting data for fusion, defining a system-defined cleaning entry, and actively cleaning data according to a time period or a unique identifier are as follows:

[0046] Use ETL tools to fully extract the initialized approval process data. If the approval process data is not initialized, obtain heterogeneous data from different business systems and register them in the heterogeneous system pool. After data source adaptation, extract the heterogeneous data in batches according to conditions and put them into the data extraction pool.

[0047] If there is data in the data extraction pool, the heterogeneous data or the approval process data is format-standardized, and the null values, abnormal values ​​and duplicate values ​​are cleaned and put into a unified data model conversion center;

[0048] The model conversion center maps fields of different data extraction methods through a mapping table in a database, configures an API service-type extraction method, and writes a program template to template the script tool program;

[0049] Through the model conversion center, the heterogeneous data and the approval process data are matched with each other in fields, unified in data coding, and updated;

[0050] If the approval process is not completed, the heterogeneous data and the approval process data will be placed in an unfinished data pool. If the automatic cleaning task time arrives, the system will automatically extract, clean and merge. Otherwise, the user will actively extract, clean and merge and extract data according to the data identifier and place it in the data extraction pool.

[0051] In a specific embodiment, the process panorama visualization module b is configured to: construct an approval process diagram according to the approval process data and update the approval process data in real time using a polling mechanism or WebSocket;

[0052] The approval process data includes approval nodes, approval status, timestamp and approver information.

[0053] In a specific embodiment, the message push module c is configured to: configure interactive message push rules, integrate machine learning models through an intelligent early warning engine, and predict delay risks in the approval process and trigger early warning signals.

[0054] In a specific embodiment, the interactive decision support module d is configured as follows: the system adjusts the scheduling strategy according to the early warning signal, and sends an OA email message reminder when the early warning mechanism is triggered.

[0055] It should be noted that the message push rules include approval node timeout threshold, push conditions and account dynamic management;

[0056] The fields of the mapping table include a business system identifier, an extraction method, an original data table field name, and a new data table field name.

[0057] It should be noted that in Figure 1 The method for process approval in the system shown is as follows Figure 2 Specifically, the method comprises the following steps:

[0058] The approval process data is input into the system. The heterogeneous data cleaning and fusion module uses the interface opened by the business system to extract the approval process data from different business systems in the form of API services, and then configures the corresponding scheduled tasks through the ETL tool, and extracts the approval process data for cleaning and fusion. The process panoramic visualization module constructs an approval process diagram based on the approval process data and displays the approval process nodes and the status of each link;

[0059] The system determines whether the current approval process is stagnant. If the current approval process is not stagnant, the system uses the machine learning model to predict whether the current approval process will be delayed. If the current approval process is stagnant or the machine learning model predicts that the current approval process will be delayed, the system triggers an early warning mechanism and sends an OA email message reminder.

[0060] If the machine learning model predicts that the approval will not be delayed, the priority of the current approval process is calculated based on the load of the current approvers and the approval process data sorted by priority is generated, and then the approval is performed according to the priority of the approval process data.

[0061] It should be noted that the early warning mechanism collects the same type of approval process data from the business system, uses the machine learning model to predict the approval time of each approval node, and calculates the median of the approval time of each approval node as the early warning threshold, or manually sets the threshold, and monitors the current approval time of each approval node in real time. When the approval time of a certain approval node exceeds the threshold, the system triggers the early warning mechanism.

[0062] It should be noted that the ETL tool is used to extract data when the system is initialized, and the ETL tool is used to correct and repair the data when a large amount of abnormal data exists in the system.

[0063] It should be noted that ETL is the process of extracting (Extract), cleaning (Transform) and then loading (Load) the data of the business system into the data warehouse. The purpose is to integrate the scattered, non-standard and non-uniform data in the enterprise, and output a standard and unified data source to provide data basis for enterprise decision-making.

[0064] It should be noted that the incremental extraction method using the system API interface improves the extraction efficiency and reduces the load on heterogeneous systems.

[0065] It should be noted that in order to solve the problem of unified management of field mapping in different data extraction methods and the extraction process, the present invention designs a model conversion center, which is one of the innovative points of the data cleaning and fusion process of the present invention. Through mapping tables or script tools, fields can be mapped according to different systems and different extraction methods. The script tool is templated as a program, and the program template is written. The model conversion center is configured as an API service-type extraction method to adapt to special data conversion and cleaning needs.

[0066] In a specific example, for null field values, the default value of the string field is set to "unknown", or the default value of the numeric field is set to "0"; for outliers, a threshold is set, for example, the threshold is set to 1.5 times the standard deviation, abnormal data is identified and the abnormal data point information is marked in the data row notes.

[0067] It should be noted that the process panorama visualization module uses the approval process data to display a clear approval process diagram on the web page, uses standardized graphics and color coding to indicate different approval statuses, and provides interactive functions, such as using green to indicate approval, red to indicate approval rejection, and yellow to indicate pending approval;

[0068] On the approval process page, WebSocket or polling mechanism is used to implement real-time data updates to ensure that the information displayed on the user interface is up to date, and a message warning function is placed. When administrators and process initiators view the approval process, they can choose to send OA email message reminders to approvers based on the approval progress.

[0069] It should be noted that the approval process data is input into the machine learning model and the historical approval process data is used to train the machine learning model. Specifically, the format of the approval process data is converted through data preprocessing and feature engineering, and the cross-validation technology and grid search technology are used to optimize the parameters of the machine learning model.

[0070] Among them, the machine learning model includes an LSTM long short-term memory network model and an XGBoost regression tree enhancement model.

[0071] It should be noted that the LSTM long short-term memory network model monitors the validation set during training. If the loss of the validation set does not decrease as the training progresses, or the training set is overfitted, the parameters of the LSTM long short-term memory network model are modified to optimize the model configuration.

[0072] Modifying the parameters of the LSTM long short-term memory network model includes reducing the learning rate, and increasing the batch size, hidden layer neurons, number of iterations, and dropout rate.

[0073] It should be noted that the XGBoost regression tree boosting model improves the verification accuracy by reducing the learning rate and increasing the maximum depth of the regression tree, and increases the minimum number of samples and regularization parameters for regression tree construction when the XGBoost regression tree boosting model is overfitting; further overfitting is prevented by reducing the column sampling ratio and the row sampling ratio; and

[0074] When the training time of the XGBoost regression tree enhancement model is too long, the regression tree is prevented from splitting by increasing the minimum number of samples.

[0075] It should be noted that the message push module consists of an advanced configuration function and an intelligent warning engine. The advanced configuration function is through an interactive web page, and the administrator configures the interactive message push rules through the web page. The web page adopts a responsive layout to ensure that a good user experience can be provided on different devices. In addition, the system supports multiple configuration push methods, such as multiple mailbox services, SMS gateways and enterprise WeChat;

[0076] The intelligent warning engine is the core component of the message warning center. It integrates advanced machine learning algorithms to predict potential delays in the approval process. It can automatically analyze historical data and identify bottlenecks in the approval process. When the risk of delay is predicted, it triggers a warning signal to notify relevant personnel and provide intervention suggestions to help users take measures in advance, thereby reducing the risk of delays.

[0077] It should be noted that the LSTM long short-term memory network model can capture long-term dependencies in time series data, and the XGBoost regression tree enhancement model is used to process classification and regression problems. The input features of the LSTM long short-term memory network model and the XGBoost regression tree enhancement model include the ID of the approval node, the historical approval time, the workload of the approver, and the approval time period. The features jointly determine the complexity of the approval process and the possibility of delay.

[0078] It should be noted that the LSTM long short-term memory network model and the XGBoost regression tree enhancement model use cross-validation and grid search technology to optimize model parameters during the training process to ensure the prediction accuracy and generalization ability of the LSTM long short-term memory network model and the XGBoost regression tree enhancement model at different approval nodes; the cross-validation divides the data set into multiple subsets, and uses one subset as a test set and the remaining subsets as training sets in turn to evaluate the stability and accuracy of the model.

[0079] In the specific example, the initial parameters of the LSTM long short-term memory network model are set as follows: learning rate "0.01", batch size "6464", number of hidden layer neurons "100100", number of iterations "2020", drop rate "0.30", optimizer "Adam", loss function "MSE", sequence length "100100";

[0080] Monitor the performance on the validation set during training. If the validation loss does not decrease as training progresses, or the model is overfitting on the training set, reduce the learning rate to "0.001", increase the batch size to "128128", adjust the number of hidden layer neurons to "200200", increase the number of iterations to "5050", and increase the dropout rate to "0.50" to find the best model configuration to ensure that the LSTM long short-term memory network model performs well on both the training set and the validation set.

[0081] In a specific embodiment, the initial parameters of the XGBoost regression tree enhancement model are set to: learning rate (eta) is "0.1", the maximum depth of the tree (max_depth) is "6", the minimum number of samples for tree construction (min_ch i ld_weight) is "1", the regularization parameter is set to "gamma = 1, lambda = 1", the minimum split gain for tree construction (min_sp l it_gain) is "0", the column sampling ratio (subsamp le) is "1", the row sampling ratio (col sample_bytree) is "1", and the number of early stopping rounds (early_stopping_rounds) is "8";

[0082] If the verification accuracy does not improve, first reduce the learning rate to "0.05". If the XGBoost regression tree enhancement model still performs poorly, increase the maximum depth of the tree to "10". If the model is overfitting, increase min_ch i ld_weight to "3" and increase the regularization parameter lambda to "2". To further prevent overfitting, reduce the column sampling ratio (subsamp le) to "0.8" and the row sampling ratio (col sample_bytree) to "0.7". If the model training time is too long, increase the minimum split gain (min_sp l it_gain) of tree construction to "0.01" to reduce unnecessary tree splitting. Finally, set the number of early stopping rounds (early_stopping_rounds) to "10" to achieve early stopping.

[0083] In a specific embodiment, the median of historical approval time is calculated as the threshold, and the warning threshold is set to 1.5 times the median; for example, if the median of historical approval time for a certain approval node is 2 hours, the warning threshold is 3 hours, and the system automatically triggers the warning mechanism and sends an OA email message reminder to the supervisor, including process ID, node ID, current approval time, predicted approval time and warning threshold.

[0084] It should be noted that the use of data pipelines and ETL tools to build an approval process visualization platform has greater flexibility and adaptability. The data pipeline mainly utilizes the interface open to the outside world by the business system to extract business system data in the form of API services. The ETL tool establishes the extraction logic through the function of the canvas component provided, and configures the corresponding scheduled task cycle extraction. After the data is fully and real-time cleansing and integration through the established unified model conversion center, the data enters the data center of the process visualization platform, and finally the data is constructed into a visual approval process diagram to display the status of each link of the approval in real time; the data pipeline and ETL tools integrate, clean and convert data from different business systems, solving the technical problem of the existing technology that when data is distributed in different business systems, a completely unified process engine is required. Each business system can maintain an independent process engine, and the data pipeline and ETL tools are responsible for extracting data from different systems, and after preprocessing and conversion, the data is stored uniformly.

[0085] It should be noted that the approval process data in the present invention can be effectively integrated and uniformly managed, emphasizing data consistency and centralized management, while also reducing dependence on different business system process engines. The flexibility of data pipelines and ETL tools enables enterprises to customize data processing processes according to business needs, gradually improve and optimize the approval process visualization platform, and thus better support business decision-making and process management.

[0086] It should be noted that data is extracted from multiple heterogeneous systems to form a unified data view. Through drag-and-drop visualization tools, enterprises can easily manage and integrate process tables and fields of various systems and automatically clean and convert them. In addition, users can intuitively drag and map tables and fields in different systems, define data conversion rules and loading targets; not only can data integration be easily performed, but also the consistency and accuracy of data between systems can be ensured, making the data flow of various business systems smoother.

[0087] It should be noted that the present invention focuses on the data dimension, can quickly adjust the data processing flow according to business needs, and easily integrate new systems and data sources. Through process visualization, enterprises no longer need to rely on a specific BPM process engine. Each manufacturer or system can choose the most suitable process engine according to its own business situation, enabling enterprises to quickly adapt to market changes, optimize business processes, and select the best tools according to specific needs. It has independence and flexibility and can improve overall operational efficiency.

[0088] Reference below Figure 4 , which shows a schematic diagram of the structure of a computer system suitable for implementing an electronic device of an embodiment of the present application. Figure 4 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0089] like Figure 4 As shown, the computer system includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 402 or a program loaded from a storage part 408 into a random access memory (RAM) 403. Various programs and data required for system operation are also stored in the RAM 403. The CPU 401, the ROM 402, and the RAM 403 are connected to each other via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0090] The following components are connected to the I / O interface 405: an input section 406 including a keyboard, a mouse, etc.; an output section 407 including a liquid crystal display (LCD), etc. and a speaker, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN card, a modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as needed. A removable medium 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 410 as needed, so that a computer program read therefrom is installed into the storage section 408 as needed.

[0091] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 409, and / or installed from the removable medium 411. When the computer program is executed by the central processing unit (CPU) 401, the above functions defined in the method of the present application are executed. It should be noted that the computer-readable storage medium of the present application can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, - but not limited to - a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection with one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device, or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable storage medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wireless, wireline, optical cable, RF, etc., or any suitable combination of the foregoing.

[0092] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0093] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0094] The modules involved in the embodiments of the present application may be implemented by software or by hardware.

[0095] As another aspect, the present application further provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiment; or may exist independently without being assembled into the electronic device. The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: inputs approval process data into the system, the heterogeneous data cleaning and fusion module uses the interface open to the outside of the business system to extract the approval process data in different business systems in the form of API services, and then configures the corresponding scheduled tasks through the ETL tool, and extracts the approval process data for cleaning and fusion, and the process panoramic visualization module constructs an approval process diagram according to the approval process data and displays the approval process nodes and the status of each link; the system determines whether the current approval process is stagnant, and if the current approval process is not stagnant, the machine learning model is used to predict whether the current approval process will be delayed. If the current approval process is stagnant or the machine learning model predicts that the current approval process will be delayed, the early warning mechanism is triggered and an OA email message reminder is sent; if the machine learning model predicts that the approval will not be delayed, the priority of the current approval process is calculated according to the load of the approver of the current approval and the approval process data sorted according to the priority is generated, and then the approval is performed according to the priority of the approval process data.

[0096] Finally, it should be noted that the above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above invention concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other to form a technical solution.

Claims

1. A process approval system based on active cleaning and fusion of heterogeneous data, characterized in that: It includes heterogeneous data cleaning and fusion module, process panoramic visualization module, message push module and interactive decision support module; among them, The heterogeneous data cleaning and fusion module is configured to extract and fuse data by full extraction using ETL tools, incremental extraction using the system API interface, and real-time incremental extraction using users. The system customizes the cleaning entry and actively cleans the data according to the time period or unique identifier. The specific steps of extracting data for fusion, defining a system-defined cleaning entry, and actively cleaning data according to a time period or a unique identifier are as follows: Use ETL tools to fully extract the initialized approval process data. If the approval process data is not initialized, obtain heterogeneous data from different business systems and register them in the heterogeneous system pool. After data source adaptation, extract the heterogeneous data in batches according to conditions and put them into the data extraction pool. If there is data in the data extraction pool, the heterogeneous data or the approval process data is format-standardized, and the null values, abnormal values ​​and duplicate values ​​are cleaned and put into a unified data model conversion center; The model conversion center maps fields of different data extraction methods through a mapping table in a database, configures an API service-type extraction method, and writes a program template to template the script tool program; Through the model conversion center, the heterogeneous data and the approval process data are matched with each other in fields, unified in data coding, and updated; If the approval process is not completed, the heterogeneous data and the approval process data are placed in the unfinished data pool. If the automatic cleaning task time arrives, the system automatically extracts, cleans and merges. Otherwise, the user actively extracts, cleans and merges and extracts data according to data identifiers and places them in the data extraction pool. The process panorama visualization module is configured to: construct an approval process diagram according to the approval process data and update the approval process data in real time using a polling mechanism or WebSocket; The approval process data includes approval node, approval status, timestamp and approver information; The message push module is configured to: configure interactive message push rules, integrate machine learning models through an intelligent early warning engine, and predict delay risks in the approval process and trigger early warning signals; The interactive decision support module is configured as follows: the system adjusts the scheduling strategy according to the early warning signal, and sends an OA email message reminder when the early warning mechanism is triggered.

2. The system according to claim 1, characterized in that The message push rules include approval node timeout threshold, push conditions and account dynamic management; The fields of the mapping table include a business system identifier, an extraction method, an original data table field name, and a new data table field name.

3. A method for process approval using the system as described in any one of claims 1-2, characterized in that: include: The approval process data is input into the system. The heterogeneous data cleaning and fusion module uses the interface opened by the business system to extract the approval process data in different business systems in the form of API services, and then configures the corresponding scheduled tasks through the ETL tool, and extracts the approval process data for cleaning and fusion. The process panoramic visualization module constructs an approval process diagram based on the approval process data and displays the approval process nodes and the status of each link; The system determines whether the current approval process is stagnant. If the current approval process is not stagnant, the system uses the machine learning model to predict whether the current approval process will be delayed. If the current approval process is stagnant or the machine learning model predicts that the current approval process will be delayed, the system triggers an early warning mechanism and sends an OA email message reminder. If the machine learning model predicts that the approval will not be delayed, the priority of the current approval process is calculated based on the current approver load and the approval process data sorted by priority is generated, and then the approval is performed according to the priority of the approval process data.

4. The method according to claim 3, characterized in that Input the approval process data into the machine learning model and use historical approval process data to train the machine learning model. Specifically, convert the format of the approval process data through data preprocessing and feature engineering, and use cross-validation technology and grid search technology to optimize the parameters of the machine learning model; Among them, the machine learning model includes an LSTM long short-term memory network model and an XGBoost regression tree enhancement model.

5. The method according to claim 4, characterized in that The LSTM long short-term memory network model monitors the validation set during training, and if the loss of the validation set does not decrease as the training progresses, or the training set is overfitted, the parameters of the LSTM long short-term memory network model are modified to optimize the model configuration; Modifying the parameters of the LSTM long short-term memory network model includes reducing the learning rate, and increasing the batch size, hidden layer neurons, number of iterations, and dropout rate.

6. The method according to claim 4, characterized in that The XGBoost regression tree boosting model improves the verification accuracy by reducing the learning rate and increasing the maximum depth of the regression tree, and increases the minimum number of samples and regularization parameters for regression tree construction when the XGBoost regression tree boosting model is overfitting; and prevents further overfitting by reducing the column sampling ratio and the row sampling ratio; and When the training time of the XGBoost regression tree enhancement model is too long, the regression tree is prevented from splitting by increasing the minimum number of samples.

7. A computer storable medium having one or more computer programs stored thereon, characterized in that: When the one or more computer programs are executed by a computer processor, the method according to any one of claims 3 to 6 is implemented.

Images