Order fulfillment prediction and production scheduling method and system based on customer portrait

Abstract

The application discloses an order fulfillment prediction and production scheduling method and system based on customer portrait; the scheme collects order history data, performs user portrait, marks the historical data with default, and constructs a training data set; a customer fulfillment risk prediction model is constructed according to the training data set; the fulfillment risk of each order of each customer is calculated; the dynamic running data set of order production is continuously collected, and the dynamic running data set includes an order priority sequence and a device state parameter sequence; a control instruction set is generated according to the dynamic running data set, and order distribution and device production are performed. The scheme evaluates the order fulfillment risk by performing customer portrait and customer historical order data, and adjusts the customer order production according to the fulfillment risk and the device state, thereby greatly reducing the risk loss caused by the overdue customer receivables.

Description

Technical Field

[0001] This application relates to the fields of intelligent manufacturing, industrial automation and artificial intelligence, and in particular to a method and system for order fulfillment prediction and production scheduling based on customer profiles. Background Technology

[0002] In modern manufacturing, the efficiency of production scheduling directly impacts a company's operating costs and market competitiveness. With the diversification of order demands and the increasing complexity of the production environment, traditional scheduling methods are gradually revealing problems such as slow response times and low resource utilization.

[0003] Traditional production scheduling typically relies on manual experience or a single algorithm for scheduling. When faced with dynamically changing production demands, this approach often requires a long adjustment period and can easily lead to problems such as equipment idleness or logistics bottlenecks.

[0004] In addition to considerations such as response speed and resource utilization, it is also necessary to consider the customer's performance risk. If the customer fails to pay the order fee on time, it will not only affect the production of the current order, but also affect other subsequent orders.

[0005] The current production scheduling system fails to fully consider the synergy between orders, equipment, and fulfillment risks, resulting in limited improvements in overall production efficiency. Furthermore, the existing system lacks the ability to make flexible and dynamic adjustments based on fulfillment risks. Summary of the Invention

[0006] The main purpose of this application is to provide a method and system for order fulfillment prediction and production scheduling based on customer profiles, which can improve the response speed and resource utilization of production scheduling in the manufacturing industry.

[0007] To achieve the above objectives, embodiments of the present invention provide a method for order fulfillment forecasting and production scheduling based on customer profiles, the method comprising the following steps: Collect historical order data, create user profiles, label historical data with defaults, and build a training dataset; build a customer performance risk prediction model based on the training dataset; calculate the performance risk of each customer and each order. The system continuously collects dynamic operational datasets for order production, including order priority sequences and equipment status parameter sequences; it then generates control instruction sets based on these datasets to execute order allocation and equipment production.

[0008] Accordingly, embodiments of this application also provide an order fulfillment prediction and production scheduling system based on customer profiles, the system comprising: The customer performance risk prediction system is used to collect historical order data, create user profiles, label historical data with defaults, and build a training dataset; based on the training dataset, a customer performance risk prediction model is built; and the performance risk of each customer and each order is calculated. The intelligent agent system is used to continuously collect dynamic operation datasets of the production area. The dynamic operation datasets include order priority sequences, equipment status parameter sequences, and logistics route planning sequences. Based on the dynamic operation datasets, a multi-level linkage control instruction set is generated to execute order allocation and equipment production.

[0009] In summary, by adopting the technical solution of this application, customer profiling is conducted to assess the customer's performance risk at different stages, and the production schedule of customer orders is adjusted according to the performance risk, which greatly reduces the risk of loss caused by overdue accounts receivable. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of a customer profile-based order fulfillment prediction and production scheduling system in an embodiment of this application; Figure 2 A flowchart illustrating the order fulfillment prediction and production scheduling method based on customer profiles provided in this application embodiment; Figure 3 This is a schematic diagram illustrating the process of generating a control instruction set based on the dynamic operation dataset and executing order allocation and equipment production, provided in an embodiment of this application. Figure 4 A flowchart illustrating the process of performing multi-level feature extraction on the dynamic running dataset to generate a comprehensive scheduling evaluation matrix for each production area, as provided in this embodiment of the application. Figure 5 This is a flowchart illustrating how the scheduling scheme is classified according to a preset production efficiency optimization model to obtain scheduling instructions corresponding to the production scheme categories, as provided in this embodiment of the application. Figure 6 A schematic diagram of the structure of a customer profile-based order fulfillment prediction and production scheduling system provided in an embodiment of this application; Figure 7 A schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0013] This application provides a method and system for order fulfillment prediction and production scheduling based on customer profiles, which will be described in detail below.

[0014] like Figure 1 As shown, an order fulfillment prediction and production scheduling system based on customer profiles is provided. The system may include a customer fulfillment risk prediction system and at least one intelligent agent deployed in the production area, including an order intelligent agent, an equipment intelligent agent, and a production scheduling device.

[0015] The customer performance risk prediction system is used to collect historical order data, create user profiles, label historical data with defaults, and build a training dataset; build a customer performance risk prediction model based on the training dataset; calculate the performance risk of each customer and each order, and establish an order priority sequence.

[0016] The order agent and the equipment agent work together, exchanging information and coordinating to achieve dynamic production scheduling. First, the agents continuously collect dynamic operational datasets from their respective domains, including order priority sequences and equipment status parameter sequences.

[0017] The order intelligence agent constantly monitors the addition, modification, or cancellation of orders to ensure that the order priority sequence is updated in real time.

[0018] The device intelligence agent integrates the data acquired by the device's sensors into a sequence of device status parameters in order to fully understand the device's operating status.

[0019] The production scheduling equipment is used to continuously collect dynamic operation datasets of production areas through multiple intelligent agents in the intelligent agent system; perform multi-level feature extraction processing on the dynamic operation datasets to generate a comprehensive scheduling evaluation matrix for each production area; classify the production scheduling schemes of the comprehensive scheduling evaluation matrix according to a preset production efficiency optimization model to obtain scheduling instruction codes and resource allocation codes corresponding to the production scheduling scheme categories; generate a multi-level linkage control instruction set based on the scheduling instruction codes and resource allocation codes, the multi-level linkage control instruction set including order priority adjustment instructions and equipment start / stop control instructions; and execute equipment start / stop operations and order allocation operations according to the instruction priorities.

[0020] refer to Figure 2 , Figure 2This is a flowchart illustrating a customer profile-based order fulfillment prediction and production scheduling method provided in this application embodiment. The execution entity of this method can be a computer device, which can be a single computer device or a cluster of multiple computer devices. The computer device can be a terminal device or a server, etc. The customer profile-based order fulfillment prediction and production scheduling method provided in this application embodiment specifically includes: Step S202: Collect historical order data and create user profiles; In some embodiments, order history data includes various data such as historical data of the entire order scheduling process within a preset time period, user-related legal proceedings, business operations, and breach of contract information. The breach of contract information may include the time of breach, the amount of the breach, and the number of breaches.

[0021] In some embodiments, historical data is preprocessed, including handling missing values ​​and outliers; features are extracted from historical data, including statistical features, discretization features, time-period trend features, and profiling features. Profiling features include company size, company type, and company default history.

[0022] Step S204: Label the historical data with defaults to construct a training dataset; In some embodiments, correlation analysis is performed on the extracted features to identify one or more features that are more correlated with the default event than a preset threshold, which are then used as training samples.

[0023] Step S206: Construct a customer performance risk prediction model based on the training dataset; In some embodiments, a customer performance risk prediction model is constructed based on the XGBoost algorithm.

[0024] First, a set of hyperparameters is selected, including `colsample_bytree` (column sampling ratio), `eta` (weights at each step, similar to the learning rate), `gamma` (minimum loss reduction required when splitting leaf nodes), `max_depth` (tree depth), `min_child_weight` (minimum leaf node weight), `n_estimators` (number of trees), `reg_alpha` (L1 regularization coefficient for weights), `reg_lambda` (L2 regularization coefficient for weights), `scale_pos_weight` (positive and negative weight balance), and `subsample` (subsampling ratio). After obtaining the hyperparameter combinations, they are substituted into the XGBoost model to begin training. Model evaluation metrics are constructed, Bayesian optimization is used to adjust model parameters to obtain optimal performance, and Hyperopt is used to automate hyperparameter tuning, saving parameter selection time and resources.

[0025] Step S208: Calculate the current fulfillment risk for each customer and each order; In some embodiments, the data of each customer's order are preprocessed and feature extracted, and one or more features that are more correlated with default events than a preset threshold are selected as inputs to the customer performance risk prediction model to obtain the corresponding performance risk.

[0026] Step S210: Continuously collect dynamic operation datasets for order production, wherein the dynamic operation datasets include order priority sequences and equipment status parameter sequences; In some embodiments, the order priority sequence is a data sequence generated by an order agent based on customer order information and order fulfillment risk. Customer order information includes numerous factors, such as the order's delivery date requirements, the importance of the products involved, and any special customization requirements. The order agent comprehensively evaluates the customer order information and order fulfillment risk according to specific rules and algorithms to determine the priority of each order. For example, for an automobile manufacturer, if a customer orders a high-end car with special configurations and a very tight delivery date, and the order fulfillment risk is low, this order will be marked as high priority by the order agent and placed at the top of the order priority sequence.

[0027] The equipment status parameter sequence is a collection of data about the status of production line equipment collected by the equipment agent. The production line equipment status data covers various parameters during equipment operation, such as equipment operating speed, equipment temperature, equipment load rate, and the wear degree of key equipment components. These data are arranged in a certain time order or logical order to form the equipment status parameter sequence.

[0028] In this step, the order agent and the equipment agent continuously collect relevant data within the production area. The order agent constantly receives new order information and order fulfillment risk information, and updates the order priority sequence in real time; the equipment agent continuously collects equipment status parameters using various sensors installed on the equipment (such as temperature sensors, pressure sensors, speed sensors, etc.). This continuous collection of data provides a comprehensive and real-time data foundation for subsequent production scheduling decisions. By continuously collecting these dynamic operational datasets, the production system can promptly understand various changes in the production process and avoid unreasonable production scheduling due to data lag.

[0029] Step S212: Generate a control instruction set based on the dynamic operation dataset, and execute order allocation and equipment production.

[0030] In some embodiments, as shown in the accompanying drawings Figure 3 As shown, generating a control instruction set based on the dynamic operation dataset and executing order allocation and equipment production includes the following sub-steps: Step S302: Perform multi-level feature extraction processing on the dynamic operation dataset to generate a comprehensive scheduling evaluation matrix for each production area; wherein, the comprehensive scheduling evaluation matrix includes an order completion probability vector and an equipment load distribution vector.

[0031] Multi-level feature extraction is a method for in-depth analysis of different types of data to uncover valuable features. For example, time series analysis might be used for order priority sequences with time-series characteristics; spatial feature analysis might be used for equipment status parameter sequences, which contain spatial structure information. Through analysis at different levels, comprehensive feature information is extracted from the data.

[0032] The order completion probability vector is a vector representation obtained by analyzing relevant data such as order priority sequences. It reflects the likelihood that each order can be completed on time under the current production environment.

[0033] The equipment load distribution vector is a vector derived from the analysis of equipment state parameter sequences. It describes the load status of each piece of equipment on the production line. Each element of the vector corresponds to a piece of equipment, and the value of the element indicates the load level of that equipment.

[0034] The comprehensive scheduling evaluation matrix is ​​a matrix formed by combining order completion probability vectors and equipment load distribution vectors according to certain rules. This matrix comprehensively reflects the overall status of orders and equipment within the production area, providing comprehensive data for subsequent production scheduling scheme classification.

[0035] In one embodiment, data mining and machine learning techniques can be used to achieve multi-level feature extraction. For the order priority sequence, a Long Short-Term Memory (LSTM) network is used to process its time-series features. The LSTM network can learn the changing patterns of order priorities over time, predict the order completion probability based on historical order data and the current order status, and thus construct an order completion probability vector. For the equipment status parameter sequence, a Convolutional Neural Network (CNN) is used for spatial feature extraction. Treating the equipment status parameters as two-dimensional spatial data (e.g., viewing the parameters of different components of the equipment as different points in space), the CNN can automatically extract key patterns in the equipment status, such as concentrated load areas and potential fault areas, thereby generating an equipment load distribution vector. Finally, this vector is concatenated according to the production area number to form a comprehensive scheduling evaluation matrix.

[0036] Step S304: Based on the preset production efficiency optimization model, classify the production scheduling schemes of the comprehensive scheduling evaluation matrix to obtain the scheduling instructions corresponding to the production scheduling scheme categories.

[0037] In this application, the production efficiency optimization model is a pre-built and trained model used to classify production scheduling schemes based on information in the comprehensive scheduling evaluation matrix. This model is trained using a large amount of historical and simulated data, based on various patterns and constraints in the production process, such as equipment production capacity and order delivery requirements.

[0038] Production scheduling scheme classification refers to inputting a comprehensive scheduling evaluation matrix into a production efficiency optimization model. The model then classifies the production scheduling schemes based on information such as the order completion probability vector, equipment load distribution vector, and logistics transportation efficiency vector within the matrix. For example, production scheduling schemes can be categorized into different types such as urgent order priority processing, equipment load balancing, and logistics optimization.

[0039] Each production scheduling scheme category corresponds to one or more scheduling instructions.

[0040] In this step, the production efficiency optimization model receives the comprehensive scheduling evaluation matrix as input. Based on its internal algorithm and pre-trained knowledge, the model performs a comprehensive analysis of the order completion probability vector and equipment load distribution vector within the comprehensive scheduling evaluation matrix. In this way, the production efficiency optimization model can automatically generate suitable production scheduling plans based on the current production status, improving the intelligence level of scheduling, reducing the subjectivity and errors that may result from manual intervention, and simultaneously improving the rationality of resource allocation and production efficiency.

[0041] As one possible implementation, the production efficiency optimization model can be built based on reinforcement learning algorithms. First, a large amount of historical production data is collected, including different comprehensive scheduling evaluation matrices and their corresponding optimal production scheduling schemes. Then, this data is used as training samples to construct a reinforcement learning environment. In the reinforcement learning environment, the agent (i.e., the production efficiency optimization model) takes action based on the current state of the comprehensive scheduling evaluation matrix (i.e., selects a production scheduling scheme category), and the environment provides rewards (e.g., a positive reward if the selected production scheduling scheme improves production efficiency; a negative reward if it leads to production delays or resource waste). Through continuous interaction with the environment, the production efficiency optimization model learns the optimal production scheduling scheme classification strategy. In practical applications, when a new comprehensive scheduling evaluation matrix is ​​received, the model outputs the production scheduling scheme category according to the learned strategy and looks up the corresponding scheduling instruction code and resource allocation code according to a predefined mapping relationship.

[0042] The scheduling instructions include order priority adjustment instructions and equipment start / stop control instructions.

[0043] The order priority adjustment instruction is used to adjust the priority of orders. When new situations arise during the production process (such as the insertion of urgent orders, delayed delivery of certain orders, etc.), it is necessary to adjust the order priorities. This instruction contains information such as the target order number and the priority adjustment value.

[0044] Equipment start / stop control commands are used to control the start and stop of equipment on the production line. This command includes information such as the target equipment number and start / stop status flags.

[0045] In one embodiment, reference Figure 4 Step S302: Performing multi-level feature extraction processing on the dynamic running dataset to generate a comprehensive scheduling evaluation matrix for each production area may specifically include the following steps: Step S402: Perform trend analysis on the order priority sequence to construct an order completion probability vector.

[0046] Trend analysis is used to analyze the changes in order priority sequences over time and other factors. It not only focuses on the current status of order priorities but also considers their changes over a past period and potential future trends.

[0047] The order completion probability vector is a vector specifically constructed based on trend analysis of the order priority sequence. It represents the probability of completing each order after considering its priority trend. Each element in the vector corresponds to an order, and the value of the element reflects the likelihood that the order can be completed on time under the current production trend. For example, for an order that consistently has a high priority and no other interfering factors, its corresponding element value in the order completion probability vector will be relatively high, indicating that the order has a greater probability of being completed on time.

[0048] This step focuses on trend analysis of order priority sequences to construct order completion probability vectors. First, order priority sequence data needs to be collected over a certain time range. This time range should be long enough to reflect the changing trends in order priorities, but not so long as the data becomes outdated. Then, the changing patterns in this data are analyzed. By quantitatively analyzing various trends in the order priority sequences, the completion probability of each order is determined, thus constructing an order completion probability vector. This trend-based order completion probability vector more accurately reflects the likelihood of order completion in the actual production process, providing more valuable information for subsequent production scheduling decisions and avoiding inaccurate scheduling arrangements based solely on current priorities.

[0049] One possible implementation is to use trend fitting methods from time series analysis to perform trend analysis on order priority sequences. First, the order priority sequence is treated as time series data. Then, a suitable trend fitting model, such as a multinomial fitting model, is selected. For each order's priority sequence, a multinomial fitting model is used to fit its trend over time. Based on the fitted multinomial function, the priority trend of the order over a future period can be predicted. Next, based on the priority trend, current production resource status, and competitive relationships between orders, a probability calculation model is used to calculate the completion probability of each order. For example, a function based on priority trend, resource competition, and order complexity can be defined, and the quantified values ​​of these factors can be substituted into the function to obtain the completion probability of each order, thereby constructing an order completion probability vector.

[0050] In one embodiment, the order priority sequence is processed by trend analysis based on a long short-term memory network to obtain order priority change trend information; and an order completion probability vector is constructed based on the order priority change trend information.

[0051] Step S404: Extract spatial features from the device state parameter sequence and generate a device load distribution vector based on the extracted spatial features.

[0052] In this application, spatial feature extraction is an analysis method for equipment state parameter sequences, which treats equipment state parameters as data with spatial structure. This spatial structure can be the physical spatial relationship between various components within the equipment, or the relative spatial layout of the equipment with other equipment on the production line.

[0053] The equipment load distribution vector is a vector generated based on the spatial feature extraction results of the equipment state parameter sequence. It reflects the load distribution of the equipment on its spatial structure. Each element in the vector corresponds to a part or region in the equipment's spatial structure, and the value of the element represents the load level of that part or region.

[0054] In one embodiment, spatial features can be extracted from the device state parameters based on a convolutional neural network to obtain spatial features; key pattern recognition results of the device can be determined based on the spatial features, and a device load distribution vector can be generated based on the key pattern recognition results.

[0055] Step S406: The order completion probability vector and the equipment load distribution vector are concatenated according to the production area number to form a comprehensive scheduling evaluation matrix.

[0056] In this application, splicing refers to the operation of combining the order completion probability vector and the equipment load distribution vector according to a specific rule. This specific rule involves arranging the vectors according to production area numbers, so that the elements in the vectors correspond to the same production area, thus forming a comprehensive matrix structure.

[0057] This step begins by determining the production area numbers. These numbers are identifiers used to distinguish different production areas, each with its own unique order, equipment, and logistics information. Then, the order completion probability vector, equipment load distribution vector, and logistics efficiency vector are arranged according to the production area numbers. For example, if there are three production areas, numbered 1, 2, and 3, the elements in the order completion probability vector will be arranged in the order of production area 1, 2, and 3; the equipment load distribution vector will also be arranged according to the production area numbers. Arranging the vectors according to this rule and combining them forms the comprehensive scheduling evaluation matrix. This matrix allows the previously scattered order and equipment information to be represented in a unified structure, facilitating analysis by the production efficiency optimization model. Through the comprehensive scheduling evaluation matrix, the production system can more intuitively understand the overall situation of different production areas, thereby making more scientific production scheduling decisions and avoiding decision-making errors caused by information dispersion.

[0058] In one embodiment, concatenating the order completion probability vector and the equipment load distribution vector according to the production area number to form a comprehensive scheduling evaluation matrix may specifically include: Priority percentage data for each time period is extracted from the order completion probability vector, and a normalized order completion probability vector is generated by weighted averaging over the time window.

[0059] Key load index data for each device are extracted from the device load distribution vector, and standardized device load distribution vectors are generated by clustering by device type.

[0060] The normalized order completion probability vector and the standardized equipment load distribution vector are arranged and concatenated according to preset dimensions to generate a comprehensive scheduling evaluation matrix.

[0061] In this application, the preset dimension arrangement refers to determining the order and dimension of the normalized order completion probability vector and the standardized equipment load distribution vector in the comprehensive scheduling evaluation matrix according to preset rules.

[0062] In one embodiment, reference Figure 5 Step S304: Classify the production scheduling schemes of the comprehensive scheduling evaluation matrix according to the preset production efficiency optimization model to obtain scheduling instructions corresponding to the production scheduling scheme categories. These instructions may specifically include: Step S502: Input the comprehensive scheduling evaluation matrix into the production efficiency optimization model trained by the reinforcement learning algorithm, so as to output the optimal production scheduling scheme category by simulating different production scenarios.

[0063] The production efficiency optimization model is built upon reinforcement learning algorithms and aims to find the optimal production scheduling scheme category based on information from the comprehensive scheduling evaluation matrix. By learning from a large amount of production scenario data, this model can understand the complex relationships between different orders and equipment conditions (elements in the comprehensive scheduling evaluation matrix), and the impact of these relationships on production efficiency. For example, the model can learn how to adjust the production scheduling scheme to improve overall production efficiency when equipment load is high but order priority is low.

[0064] The optimal production scheduling scheme category is the result output by the production efficiency optimization model based on the comprehensive scheduling evaluation matrix and reinforcement learning algorithm.

[0065] In one embodiment, the comprehensive scheduling evaluation matrix is ​​preprocessed, including data standardization and normalization. A reinforcement learning environment is constructed based on the preprocessed comprehensive scheduling evaluation matrix, including a state space, an action space, and a reward mechanism. Based on the reinforcement learning environment and the production efficiency optimization model, the optimal production scheduling scheme category is output by simulating different production scenarios.

[0066] Based on a reinforcement learning environment and a production efficiency optimization model, the model outputs the optimal production scheduling scheme category by simulating different production scenarios. In the constructed reinforcement learning environment, the production efficiency optimization model begins simulating various production scenarios. According to the reinforcement learning algorithm, the model tries different actions (i.e., different production scheduling schemes) in each scenario and receives corresponding rewards or penalties based on the reward mechanism. Through continuous learning and exploration in these scenarios, the model gradually finds the production scheduling scheme category that achieves optimal production efficiency under various conditions. For example, if the simulation reveals that prioritizing urgent orders and rationally arranging equipment and logistics can achieve the highest production efficiency in most scenarios, then "urgent order priority" may be identified as the optimal production scheduling scheme category.

[0067] Step S504: Map the optimal production scheduling scheme category to a preset scheduling instruction.

[0068] Scheduling instruction codes can be recognized by the central scheduling platform and corresponding operations can be executed. For example, a scheduling instruction might indicate operations such as starting a specific piece of equipment, adjusting the priority of an order, or changing a logistics route. Each scheduling instruction is associated with a specific production scheduling scheme category to ensure that production is scheduled according to the optimal scheduling scheme during the production process.

[0069] Accordingly, to better implement the above methods, embodiments of this application also provide an order fulfillment prediction and production scheduling system based on customer profiles, such as... Figure 6 As shown, the system includes: The Customer Performance Risk Prediction System 602 is used to collect historical order data, create user profiles, label historical data with defaults, and build a training dataset; build a customer performance risk prediction model based on the training dataset; and calculate the performance risk of each customer and each order. The intelligent agent system 604 is used to continuously collect dynamic operation datasets of the production area. The dynamic operation datasets include order priority sequences, equipment status parameter sequences, and logistics path planning sequences. Based on the dynamic operation datasets, a multi-level linkage control instruction set is generated to execute order allocation and equipment production.

[0070] The implementation details of each module are provided in the preceding method embodiments and will not be repeated here. The technical effects achieved by each module and device are described in the foregoing method embodiments.

[0071] It should be noted that, in practical implementation, the above modules can be arbitrarily combined and integrated into one or more modules, or implemented as independent entities. Furthermore, the above modules can be implemented in hardware or as software functional modules. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. The aforementioned storage medium can be a read-only memory, a hard disk, or an optical disk, etc.

[0072] like Figure 7 As shown, this application embodiment also provides a computer device 70, characterized in that it includes a processor 701 and a memory 702, wherein the memory 702 stores a computer program, and when the computer program is executed by the processor 701, the processor 701 performs the steps of any of the methods described above.

[0073] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0074] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0075] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. A method for order fulfillment forecasting and production scheduling based on customer profiles, characterized in that, The method includes the following steps: Collect historical order data, create user profiles, label historical data with defaults, and build a training dataset; build a customer performance risk prediction model based on the training dataset; calculate the performance risk of each customer and each order. The system continuously collects dynamic operational datasets for order production, including order priority sequences and equipment status parameter sequences; it then generates control instruction sets based on these datasets to execute order allocation and equipment production.

2. The method according to claim 1, characterized in that, The customer profiling system collects historical order data, including data on the entire order process and customer-related legal proceedings. To create user profiles, historical data showing defaults are labeled, and a training dataset is constructed, including: Extract user data to create user profiles and extract features from historical data; Historical data are labeled based on the amount of default and the time of default to construct a data training set.

3. The method according to claim 2, characterized in that, The customer performance risk prediction model is the XGBoost model.

4. The method according to claim 1, characterized in that, The training set includes: The sample consists of order information, user profiles, and corresponding labels indicating whether a breach of contract has occurred.

5. The method according to claim 1, characterized in that, The order priority sequence is generated based on performance risk and order information.

6. The method according to claim 1, characterized in that, Generating a set of control instructions based on the dynamic runtime dataset includes: Multi-level feature extraction processing is performed on the dynamic operation dataset to generate a comprehensive scheduling evaluation matrix for each production area; wherein, the comprehensive scheduling evaluation matrix includes an order completion probability vector and an equipment load distribution vector; Based on the preset production efficiency optimization model, the comprehensive scheduling evaluation matrix is ​​processed to classify the production scheduling schemes, and scheduling instructions corresponding to the production scheduling scheme categories are obtained. A control instruction set is generated based on the scheduling instructions, and the control instruction set includes order priority adjustment instructions and equipment start / stop control instructions.

7. The method according to claim 6, characterized in that, Multi-level feature extraction processing is performed on the dynamic running dataset to generate a comprehensive scheduling evaluation matrix for each production area, including: The order priority sequence is subjected to trend analysis to construct an order completion probability vector; Spatial features are extracted from the sequence of equipment state parameters, and a equipment load distribution vector is generated based on the extracted spatial features; The order completion probability vector and the equipment load distribution vector are concatenated according to the production area number to form a comprehensive scheduling evaluation matrix.

8. The method according to claim 7, characterized in that, The order priority sequence is subjected to trend analysis to construct an order completion probability vector, including: The order priority sequence is processed by a long short-term memory network to obtain the order priority change trend information; an order completion probability vector is constructed based on the order priority change trend information. Spatial feature extraction is performed on the device state parameter sequence, and a device load distribution vector is generated based on the extracted spatial features. This includes extracting spatial features from the device state parameters based on a convolutional neural network to obtain spatial features; determining the key pattern recognition result of the device based on the spatial features; and generating the device load distribution vector based on the key pattern recognition result.

9. The method according to claim 6, characterized in that, The comprehensive scheduling evaluation matrix is ​​classified according to a preset production efficiency optimization model to obtain scheduling instructions corresponding to the production scheduling scheme categories, including: The comprehensive scheduling evaluation matrix is ​​input into the production efficiency optimization model trained by the reinforcement learning algorithm, so as to output the optimal production scheduling scheme category by simulating different production scenarios; Based on the optimal production scheduling scheme category, a scheduling instruction corresponding to the production scheduling scheme category is generated.

10. An order fulfillment prediction and production scheduling system based on customer profiles, characterized in that, The system includes: The customer performance risk prediction system is used to collect historical order data, create user profiles, label historical data with defaults, and build a training dataset; based on the training dataset, a customer performance risk prediction model is built; and the performance risk of each customer and each order is calculated. The intelligent agent system is used to continuously collect dynamic operation datasets of the production area. The dynamic operation datasets include order priority sequences, equipment status parameter sequences, and logistics route planning sequences. Based on the dynamic operation datasets, a multi-level linkage control instruction set is generated to execute order allocation and equipment production.

Images