Order matching method and system based on machine learning

Abstract

The invention discloses an order matching method and system based on machine learning. The method comprises the steps of obtaining a maintenance order and maintainer data; the method comprises the following steps: generating geographically adjacent and skill-compatible order sub-clusters based on a geographic position and a maintenance type, predicting the lower limit of the number of required maintainers for each sub-cluster according to a maintenance order, initializing a distribution pool according to the lower limit, constructing an order-maintainer bipartite graph in the sub-cluster, generating a joint embedded vector by applying a graph neural network, and constructing an order-maintainer bipartite graph. A part of order distribution is randomly removed in each round, and reinsertion is carried out after skill matching and man-hour constraint are forcibly met in the repair stage; and if the constraint conflicts, dynamically expanding the maintenance worker distribution pool, iterating until the scheme converges, and outputting a final order distribution scheme. According to the invention, the space efficiency of order matching and the rationality and accuracy of resource allocation are improved, the constraint compatibility and adaptability are guaranteed, and the service efficiency and satisfaction are optimized.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an order matching method and system based on machine learning. Background Art

[0002] In the field of repair services, the efficiency of matching orders with repair workers directly impacts service response speed, resource utilization, and user satisfaction. Existing order matching often suffers from irrational spatial scheduling, failing to fully consider the geographical distribution of orders. This results in excessive travel time for repair workers across regions, reducing service efficiency. When it comes to worker allocation, estimates often rely on a single metric (such as order quantity or estimated labor hours), ignoring the differentiated skill requirements of diverse repair types, which can lead to staff shortages or redundancies. Furthermore, a lack of comprehensive consideration of skill matching and labor hour allocation often results in substandard service quality due to incomplete skill coverage, or service progress is impacted by an imbalance in labor hour load, thus affecting the overall stability and reliability of the service. Summary of the Invention

[0003] In order to solve at least one of the technical problems raised above, the present invention provides an order matching method and system based on machine learning.

[0004] In a first aspect, the present invention provides an order matching method based on machine learning, the method comprising: Obtaining order data and maintenance worker data, wherein the order data includes geographic location, maintenance type set, and estimated maintenance duration, and the maintenance worker data includes skill set and maximum daily working hours; Based on the geographical location of the orders, order subclusters are generated using a density clustering algorithm; Based on the total number of orders, total estimated working hours, and skill diversity index within the order sub-cluster, the lower limit of the number of maintenance workers required for each order sub-cluster is predicted; Create a maintenance worker allocation pool with the lower limit of the number of maintenance workers as the initial value; Construct a bipartite graph of orders and repairmen within the sub-cluster, and generate a joint embedding vector integrating location, repair type, and duration through a graph neural network. Based on the joint embedding vector, multiple rounds of damage repair cycles are executed, and skill matching and work time constraints are enforced in each repair phase. When the constraints cannot be met, the repairman allocation pool is dynamically expanded until convergence conditions are reached; and the final order allocation plan is output.

[0005] Preferably, the method of predicting the lower limit of the number of maintenance workers required for each order subcluster based on the total number of orders, total estimated working hours, and skill diversity index in the order subcluster includes: Calculate the ratio of the total estimated working hours for orders in the sub-cluster to the average daily working hours of maintenance workers as the base quantity; The base number is weighted and modified according to the skill diversity index, where the skill diversity index is measured by the uniformity of the distribution of skill types; Divide the total number of orders by the average daily processing capacity of maintenance workers as a reference value; The maximum value between the weighted adjusted basic quantity and the quantity reference value shall be taken as the lower limit of the number of maintenance workers.

[0006] Preferably, the step of constructing a bipartite graph of orders and maintenance workers within a sub-cluster and generating a joint embedding vector integrating location, maintenance type, and duration through a graph neural network includes: All orders in the sub-cluster and the repairmen in the repairmen allocation pool are treated as independent node sets of a bipartite graph, and fully connected edges are established between order nodes and repairmen nodes; Assign a multidimensional feature vector to each order node, including the normalized geographic location coordinates, the multi-hot encoding vector of the maintenance type set, and the deviation value between the expected maintenance time and the standard time; Each maintenance worker node is assigned a multi-dimensional feature vector, including the coordinates of the historical service area center, the overlap vector between the skill set and the maintenance type requirements of the current sub-cluster, and an efficiency factor calculated based on historical work orders; A three-layer graph attention neural network is used to process the bipartite graph. The geographical location correlation, maintenance type matching and time adaptability features are aggregated through message passing between nodes. The final output is a joint embedding vector with a dimension of 128, where each embedding vector represents the comprehensive matching degree of a specific order-repairman pairing in terms of spatial accessibility, skill coverage adequacy and work load balance.

[0007] Preferably, the method executes multiple rounds of damage repair cycles based on the joint embedding vector, forcing skill matching and work hour constraints to be met in each repair phase, and dynamically expanding the repair worker allocation pool when the constraints cannot be met until convergence conditions are reached, including: Use weighted probability to remove a preset proportion of existing assignments, where pairs with skill matching below a threshold or work load exceeding the limit are removed first; Reassign maintenance workers to the removed orders based on reallocation constraints, including that the maintenance worker's skill set must include all repair types for the order and that the total duration of a single maintenance worker's assigned orders must be less than the product of the maximum daily working time and the utilization threshold. When the number of consecutive failures in the repair phase reaches the preset upper limit, based on the missing skill types and the amount of work hour gap, candidate repairmen with skill coverage of more than 90% and historical work hour utilization of less than 85% will be transferred from surrounding areas to join the repairman allocation pool; When the global matching degree improvement in three consecutive cycles is less than 0.5%, it is determined that the convergence condition has been met.

[0008] In a second aspect, the present invention further provides an order matching system based on machine learning, the system comprising: A data acquisition module is used to acquire order data and maintenance worker data. The order data includes geographic location, maintenance type set, and estimated maintenance duration. The maintenance worker data includes skill set and maximum daily working hours. The order clustering module is used to generate order subclusters based on the geographical location of the order through a density clustering algorithm; Maintenance worker demand prediction module, which is used to predict the lower limit of the number of maintenance workers required for each order subcluster based on the total number of orders, total estimated working hours and skill diversity index in the order subcluster; The maintenance worker allocation pool initialization module is used to create a maintenance worker allocation pool with the lower limit of the number of maintenance workers as the initial value; The graph construction and embedding module is used to construct a bipartite graph of orders and repairmen within a sub-cluster and generate a joint embedding vector that integrates location, repair type, and duration through a graph neural network. An iterative optimization matching module is configured to execute multiple rounds of damage repair cycles based on the joint embedding vector, enforce skill matching and labor time constraints in each repair phase, dynamically expand the repair worker allocation pool when the constraints cannot be met, and output a final order allocation plan.

[0009] Preferably, the maintenance worker demand forecasting module is further used to: Calculate the ratio of the total estimated working hours for orders in the sub-cluster to the average daily working hours of maintenance workers as the base quantity; The base number is weighted and modified according to the skill diversity index, where the skill diversity index is measured by the uniformity of the distribution of skill types; Divide the total number of orders by the average daily processing capacity of maintenance workers as a reference value; The maximum value between the weighted adjusted basic quantity and the quantity reference value shall be taken as the lower limit of the number of maintenance workers.

[0010] Preferably, the graph construction and embedding module is further used for: All orders in the sub-cluster and the repairmen in the repairmen allocation pool are treated as independent node sets of a bipartite graph, and fully connected edges are established between order nodes and repairmen nodes; Assign a multidimensional feature vector to each order node, including the normalized geographic location coordinates, the multi-hot encoding vector of the maintenance type set, and the deviation value between the expected maintenance time and the standard time; Each maintenance worker node is assigned a multi-dimensional feature vector, including the coordinates of the historical service area center, the overlap vector between the skill set and the maintenance type requirements of the current sub-cluster, and an efficiency factor calculated based on historical work orders; A three-layer graph attention neural network is used to process the bipartite graph. The geographical location correlation, maintenance type matching and time adaptability features are aggregated through message passing between nodes. The final output is a joint embedding vector with a dimension of 128, where each embedding vector represents the comprehensive matching degree of a specific order-repairman pairing in terms of spatial accessibility, skill coverage adequacy and work load balance.

[0011] In a third aspect, the present invention also provides an electronic device comprising a processor and a memory, wherein the memory is used to store computer program code, and the computer program code comprises computer instructions. When the processor executes the computer instructions, the electronic device executes the method as described in the first aspect above and any possible implementation thereof.

[0012] In a fourth aspect, the present invention also provides a computer-readable storage medium, in which a computer program is stored. The computer program includes program instructions, and when the program instructions are executed by a processor of an electronic device, the processor executes the method as described in the first aspect above and any possible implementation method thereof.

[0013] Compared with the prior art, the present invention has the following beneficial effects: This invention integrates spatial clustering, multidimensional resource demand prediction, graph neural network feature fusion, and a dynamic optimization loop to construct an intelligent order matching mechanism that balances spatial rationality, skill adaptability, and work-hour balance. Specifically, density clustering based on geographic location enables spatially intensive order processing, reducing the inefficiency of cross-regional scheduling. A multi-dimensional parameter approach, including total order volume, total work hours, and skill diversity, is used to predict the lower limit for the number of repair workers, providing a scientific basis for initial resource allocation. A graph neural network is then used to fuse multimodal features of the order and repair worker, such as their location, repair type, and duration, into a joint embedding vector to quantify the overall matching degree. Furthermore, a destroy-repair loop dynamically optimizes the allocation scheme, combining the mandatory satisfaction of skill and work-hour constraints with the dynamic expansion of the allocation pool to achieve continuous iterative optimization of the matching scheme, ultimately achieving efficient, accurate, and constraint-compatible order-worker matching. Density clustering is used to improve the spatial efficiency of order matching to reduce cross-regional commuting costs. The lower limit of the number of maintenance workers is predicted based on multi-dimensional parameters to optimize resource allocation and avoid redundancy or shortage. Graph neural networks are used to fuse multimodal features to improve matching accuracy. The compatibility of skill and working time constraints and the adaptability of the solution are ensured through the destruction-repair cycle and dynamic expansion of the allocation pool, ultimately comprehensively improving service response efficiency, reliability, and the mutual satisfaction of users and maintenance workers.

[0014] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the background technology, the drawings required for use in the embodiments of the present invention or the background technology will be described below.

[0016] The accompanying drawings herein are incorporated into and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and, together with the specification, are used to explain the technical solutions of the present disclosure.

[0017] Figure 1 A flowchart of an order matching method based on machine learning provided by an embodiment of the present invention; Figure 2 A schematic diagram of the structure of an order matching system based on machine learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0019] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute a separate or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0020] The current maintenance service order matching has technical problems such as unreasonable spatial scheduling, single dimension of maintenance worker quantity configuration and ignoring the need for skill diversity, lack of comprehensive consideration of skill and working time allocation, and insufficient dynamic adaptability.

[0021] See also Figure 1 , Figure 1 The following is a flow chart of an order matching method based on machine learning provided by an embodiment of the present invention. Figure 1 As shown, the method includes: S100, acquiring order data and maintenance worker data, wherein the order data includes a geographic location, a maintenance type set, and an estimated maintenance duration, and the maintenance worker data includes a skill set and a maximum daily working time; In this embodiment, in the step of obtaining order data and repairman data, the repair order set can be collected through various channels, such as repair requests submitted online by users, telephone appointment records, and demand information synchronized from third-party platforms. The geographic location information contained in each order can be specifically longitude and latitude coordinates, a detailed address, or a real-time location obtained through a positioning service. The repair type set covers the category of equipment to be repaired (such as air conditioners, refrigerators, washing machines, etc.) and the type of fault (such as refrigeration anomalies, circuit faults, mechanical damage, etc.). The estimated repair time is estimated based on the historical processing time of similar faults, the complexity of the fault description, and the experience of the repairman. The repairman set is derived from the repairman information database registered with the system. The skill set includes the equipment repair qualifications possessed by the repairman, the types of faults that can be handled, and the repair areas in which the repairman is proficient. The maximum daily working time is the total time that the repairman can pre-set to perform repair work each day, which can be deducted from necessary rest time and commuting time.

[0022] S200, based on the geographical location of the orders, generates order subclusters using a density clustering algorithm; In this embodiment, in the process of generating order subclusters based on the geographic location of the orders using a density clustering algorithm, the spatial coordinates of the geographic locations of all orders are first standardized, and the spatial distance between orders is calculated using a distance measurement method suitable for geographic space (such as actual surface distance based on latitude and longitude); then, the density clustering algorithm is used to identify dense areas of order distribution, and with each order as the core, the number of other orders within a certain spatial range around it is counted. When the number reaches a preset density threshold, these orders are grouped into the same initial cluster; the initial clusters are further merged and adjusted, and spatially adjacent initial clusters with similar density are merged into a subcluster. At the same time, for isolated sparse orders, if their distance to a subcluster is within a set range, they are included in the subcluster, ultimately forming multiple order subclusters with relatively concentrated spatial distribution.

[0023] In one possible embodiment, a neighborhood radius parameter is determined, taking into account regional traffic characteristics and repair skill distribution. The base spatial radius is determined by multiplying the target area's historical average driving speed by the average repair time for each order. Furthermore, the dispersion of the global repair type distribution is introduced as a compensation term; the more uneven the skill distribution, the greater the compensation value. During the density accessibility assessment phase, not only is the geographic distance between two orders required to be within the neighborhood radius, but the similarity between their repair type sets must also exceed a preset threshold to ensure skill compatibility among orders within the same subcluster. During the clustering process, orders that meet both the core point density requirement (a sufficient number of orders within the neighborhood) and the skill representativeness requirement (covering the most frequently occurring repair types within the neighborhood) are selected as initial core points. Density propagation is then used to form subclusters of geographically proximal orders with matching skills. Boundary points are assigned to the cluster containing the closest core point that meets the skill similarity requirement. The resulting subclusters meet the dual constraints: the geographic distance between any two orders does not exceed the regional accessibility range, and at least one repairman's skill within the cluster covers all order types.

[0024] S300, based on the total number of orders, total estimated working hours and skill diversity index in the order sub-cluster, predicts the lower limit of the number of maintenance workers required for each order sub-cluster; In this embodiment. In the step of predicting the lower limit of the number of maintenance workers based on the total number of orders, total estimated working hours and skill diversity index in the order sub-cluster, the total estimated working hours are the cumulative value of the estimated maintenance time of all orders in the sub-cluster; the skill diversity index is calculated by counting the number of types of maintenance types in the sub-cluster and the distribution of the proportion of each type. When there are many types of maintenance types and the proportion is balanced, the index is higher, and vice versa. The basic number is obtained by dividing the total estimated working hours by the average daily effective working hours of the maintenance workers (that is, the time actually used for maintenance operations during the maximum working time of the day); the quantity reference value is calculated by dividing the total number of orders by the average daily processing volume of orders by the maintenance workers (based on the average level of historical data statistics); when the basic number is weighted and corrected, the higher the skill diversity index, the greater the correction weight, so as to reflect the demand for multi-skilled maintenance workers; finally, the weighted corrected basic number is compared with the quantity reference value, and the larger value is taken as the lower limit of the number of maintenance workers required for the sub-cluster to ensure that both the total working hour demand and the order quantity scale can be met.

[0025] S400, creating a maintenance worker allocation pool with the lower limit of the number of maintenance workers as the initial value; In this embodiment, when creating a maintenance worker allocation pool with the lower limit of the number of maintenance workers as the initial value, first, based on the geographical location of the order subcluster, maintenance workers whose service areas cover the subcluster or are relatively close to it are screened out; secondly, based on the maintenance type set of the subcluster, maintenance workers whose skill sets can cover at least some maintenance types are further screened out; according to the above screening conditions, maintenance workers with a number not less than the lower limit are selected from the maintenance worker set and included in the allocation pool. If the number of maintenance workers who meet the conditions is insufficient, the maintenance workers with the highest skill matching degree and the closest distance are selected first, and the gap information is recorded; at the same time, the allocation pool will reserve a certain amount of expansion space so that maintenance workers can be dynamically supplemented according to the satisfaction of the constraints.

[0026] S500: Construct a bipartite graph of orders and repairmen within the sub-cluster and generate a joint embedding vector integrating location, repair type, and duration through a graph neural network. In this embodiment, in the process of constructing a bipartite graph of orders and repairmen within a subcluster and generating a joint embedding vector using a graph neural network, the order node set of the bipartite graph includes all orders within the subcluster, and the repairmen node set includes all repairmen within the allocation pool. Connection edges are established between order nodes and repairmen nodes to represent potential allocation relationships. In the multidimensional feature vector assigned to the order node, the normalized geographic location coordinates are obtained by mapping the geographic location of the order to a standardized numerical value within a specific spatial range. In the multi-hot encoding vector of the repair type set, each dimension corresponds to a possible repair type. If the order contains that type, the dimension value is 1, otherwise it is 0. The deviation between the estimated repair time and the standard time is calculated by the difference between the estimated order time and the average handling time for similar faults. In the multidimensional feature vector assigned to a maintenance worker node, the coordinates of the center point of the historical service area are derived from the weighted average of the geographic locations of all previous service orders for that worker. In the overlap vector between the skill set and the maintenance type requirements of the current subcluster, each dimension corresponds to a maintenance type within the subcluster, and the dimension value is a quantitative representation of whether the maintenance worker possesses the skills for that type. The efficiency factor is calculated based on the ratio of the actual processing time to the estimated processing time in the maintenance worker's historical work orders, as well as the work order completion quality score. When processing a bipartite graph using a three-layer graph attention neural network, the first layer learns a basic representation of the node's own characteristics. The second layer aggregates the geographic correlation between orders and maintenance workers through inter-node messaging (e.g., closer distances lead to higher correlation). The third layer integrates maintenance type matching (skill overlap) and time duration adaptability (the synergistic relationship between the efficiency factor and time duration deviation). The final output is a joint embedding vector that comprehensively represents the degree of match between order and maintenance worker pairs based on three dimensions: spatial accessibility, sufficient skill coverage, and work load balance.

[0027] S600: Execute multiple rounds of damage repair cycles based on the joint embedding vector, forcing skill matching and work time constraints to be met in each repair round, and dynamically expand the repair worker allocation pool when the constraints cannot be met until convergence conditions are reached; and output a final order allocation plan.

[0028] In this embodiment, multiple rounds of destruction and repair cycles are performed based on the joint embedding vector. In the destruction phase, existing assignment pairs are weighted based on the matching degree reflected by the joint embedding vector. Pairs with low matching degrees (e.g., insufficient skill matching or imbalanced work load) are assigned a higher probability of removal. A certain percentage of existing assignments are removed based on this probability. The repair phase reassigns repair workers to removed orders, strictly adhering to the following reallocation constraints: first, the repair worker's skill set must fully encompass all required repair types for the order to ensure sufficient processing capacity; second, the total estimated duration of all orders assigned to the same repair worker must not exceed the product of their maximum daily work time and a set utilization threshold to avoid work load overload. If the repair phase fails to find a repair worker assignment that meets these constraints for a removed order multiple times, the current allocation pool is deemed insufficient. Based on the missing skill types and unmet work time gaps in the sub-cluster, qualified candidate repair workers are selected from the repair worker pool in neighboring areas to be added to the allocation pool. These candidates must possess the missing key skills and have a low historical work time utilization rate to ensure sufficient free time. The global matching degree (the average matching degree of the joint embedding vectors of all order-repairman pairs) is continuously calculated during the loop. When the improvement of the global matching degree over multiple consecutive rounds is lower than the set standard, it is determined that the convergence condition has been met, the loop is stopped, and the final order allocation plan is output.

[0029] In this embodiment, a smart order matching mechanism that balances spatial rationality, skill adaptability, and work-hour balance is constructed by integrating spatial clustering, multi-dimensional resource demand prediction, graph neural network feature fusion, and a dynamic optimization loop. Specifically, density clustering based on geographic location enables spatially intensive order processing, reducing the inefficiency of cross-regional scheduling. A multi-dimensional parameter, including total order volume, total work hours, and skill diversity, is used to predict the lower limit for the number of repair workers, providing a scientific basis for initial resource allocation. A graph neural network is then used to fuse multimodal features of the order and repair worker, such as their location, repair type, and duration, into a joint embedding vector to quantify the overall matching degree. Furthermore, a destroy-repair loop dynamically optimizes the allocation scheme, combining the mandatory satisfaction of skill and work-hour constraints with the dynamic expansion of the allocation pool to achieve continuous iterative optimization of the matching scheme, ultimately achieving efficient, accurate, and constraint-compatible order-repair worker matching. Density clustering is used to improve the spatial efficiency of order matching to reduce cross-regional commuting costs. The lower limit of the number of maintenance workers is predicted based on multi-dimensional parameters to optimize resource allocation and avoid redundancy or shortage. Graph neural networks are used to fuse multimodal features to improve matching accuracy. The compatibility of skill and working time constraints and the adaptability of the solution are ensured through the destruction-repair cycle and dynamic expansion of the allocation pool, ultimately comprehensively improving service response efficiency, reliability, and the mutual satisfaction of users and maintenance workers.

[0030] Preferably, the method of predicting the lower limit of the number of maintenance workers required for each order subcluster based on the total number of orders, total estimated working hours, and skill diversity index in the order subcluster includes: Calculate the ratio of the total estimated working hours for orders in the sub-cluster to the average daily working hours of maintenance workers as the base quantity; The base number is weighted and modified according to the skill diversity index, where the skill diversity index is measured by the uniformity of the distribution of skill types; Divide the total number of orders by the average daily processing capacity of maintenance workers as a reference value; The maximum value between the weighted adjusted basic quantity and the quantity reference value shall be taken as the lower limit of the number of maintenance workers.

[0031] In this embodiment, in the process of calculating the ratio of the total estimated working hours for orders in the subcluster to the average daily working hours of maintenance workers as the basic quantity, the total estimated working hours for orders in the subcluster is the cumulative value of the estimated maintenance time corresponding to all orders contained in the subcluster, where the estimated maintenance time of each order can be comprehensively determined based on the actual processing time of historical similar maintenance orders, the complexity of the fault description and the maintenance difficulty level; the average daily working hours of maintenance workers refers to the effective time that maintenance workers can actually use to perform maintenance work within a working day, which needs to deduct necessary rest time, commuting time reserved between orders and other non-maintenance operation time. It is usually determined based on the average level of historical statistics of the average daily effective working hours of maintenance workers. Through the ratio of the above-mentioned total estimated working hours to the average daily working hours, the benchmark number of maintenance workers required to meet the working hour requirements of the subcluster is preliminarily quantified.

[0032] In the step of weighted correction of the basic quantity according to the skill diversity index, the skill diversity index is quantified by statistically analyzing the distribution characteristics of maintenance types within the sub-cluster, which may specifically include the number of maintenance types contained in the sub-cluster, the proportion of each maintenance type in the total orders, and the uniformity of distribution among different types. When there are many types of maintenance types and the difference in the proportion of each type is small, it indicates that the skill diversity index is high, and vice versa. During weighted correction, the skill diversity index is positively correlated with the correction weight, that is, the higher the skill diversity index, the greater the corresponding correction weight, so that the basic quantity is adjusted upward, thereby reflecting the additional demand for maintenance workers with multiple maintenance skills, ensuring that the corrected quantity can adapt to the diverse maintenance type needs within the sub-cluster. In the process of dividing the total number of orders by the average daily processing capacity of maintenance workers as the reference value, the total number of orders is the total number of all orders contained in the sub-cluster; the average daily processing capacity of maintenance workers refers to the average number of maintenance orders that maintenance workers can complete in one working day. This number is based on historical maintenance data statistics, taking into account factors such as the processing time of different maintenance types, the maintenance workers' operational proficiency, and the efficiency of connection between orders. It is usually taken as the average level of the average daily number of orders completed by maintenance workers in a certain period of time; through the ratio of the total number of orders to the average daily processing capacity, the reference number of maintenance workers required is quantified from the perspective of the order quantity scale, to make up for the order quantity intensity demand that may be ignored by calculating only the working hours.

[0033] In the step of taking the maximum value between the weighted-corrected basic quantity and the quantity reference value as the lower limit of the number of maintenance workers, it is necessary to compare the weighted-corrected basic quantity with the above-mentioned quantity reference value, and select the one with the larger value as the lower limit of the number of maintenance workers required for the order subcluster; this is to ensure that the lower limit can not only meet the maintenance workload requirements corresponding to the total expected working hours in the subcluster, but also cover the processing scale requirements corresponding to the total number of orders, avoid the problem of insufficient number of maintenance workers due to considering only a single dimension, provide a scientific and comprehensive quantity benchmark for the initial creation of the subsequent maintenance worker allocation pool, and ensure the rationality of the resource basis of the subsequent order matching process.

[0034] Preferably, the step of constructing a bipartite graph of orders and maintenance workers within a sub-cluster and generating a joint embedding vector integrating location, maintenance type, and duration through a graph neural network includes: All orders in the sub-cluster and the repairmen in the repairmen allocation pool are treated as independent node sets of a bipartite graph, and fully connected edges are established between order nodes and repairmen nodes; Assign a multidimensional feature vector to each order node, including the normalized geographic location coordinates, the multi-hot encoding vector of the maintenance type set, and the deviation value between the expected maintenance time and the standard time; Each maintenance worker node is assigned a multi-dimensional feature vector, including the coordinates of the historical service area center, the overlap vector between the skill set and the maintenance type requirements of the current sub-cluster, and an efficiency factor calculated based on historical work orders; A three-layer graph attention neural network is used to process the bipartite graph. The geographical location correlation, maintenance type matching and time adaptability features are aggregated through message passing between nodes. The final output is a joint embedding vector with a dimension of 128, where each embedding vector represents the comprehensive matching degree of a specific order-repairman pairing in terms of spatial accessibility, skill coverage adequacy and work load balance.

[0035] In this embodiment, all orders within the subcluster and the repairmen in the repairman allocation pool are treated as independent node sets in a bipartite graph. When establishing fully connected edges between order nodes and repairman nodes, the order node set includes all pending repair orders within the subcluster, covering repair needs of varying equipment types and fault severity. The repairman node set includes all repairmen selected from the repairman pool who preliminarily meet the basic requirements of the subcluster, covering repairmen with different skill sets and service areas. The establishment of fully connected edges aims to cover all possible order-repairman pairings. Regardless of whether the current repairman has the required skills for the order or is located nearby, the edges provide a complete potential matching space for subsequent feature learning, ensuring that the model comprehensively considers all possible pairings. In the step of assigning a multidimensional feature vector to each order node, including normalized geographic location coordinates, a multi-hot encoding vector of a maintenance type set, and a deviation value between the expected maintenance time and the standard time, the normalized geographic location coordinates convert the original geographic location information of the order (such as latitude and longitude or area code) into standardized data in a unified numerical range to eliminate the scale differences of different geographic location representation methods; in the multi-hot encoding vector of the maintenance type set, each dimension corresponds to a preset maintenance type (such as home appliances, water and electricity, etc.). If the order contains this type, the corresponding dimension in the vector is assigned a specific quantitative value, and if it does not contain this type, it is assigned another quantitative value, which intuitively presents the maintenance skill combination required for the order; the deviation value between the expected maintenance time and the standard time is calculated by comparing the expected maintenance time of the order with the average processing time of similar fault orders. A positive number indicates that the order is expected to take longer than the average level, and a negative number indicates that it is shorter than the average level, thereby reflecting the relative complexity of the order.

[0036] Each maintenance worker node is assigned a multidimensional feature vector, including the coordinates of the center point of the historical service area, the overlap vector between the skill set and the maintenance type requirements of the current sub-cluster, and an efficiency factor calculated based on historical work orders. The coordinates of the center point of the historical service area are calculated based on the geographic location of all maintenance orders completed by the maintenance worker in the past, using a weighted average (e.g., weighting by the number of services or service duration) to represent the spatial scope of the maintenance worker's primary service. In the overlap vector between the skill set and the maintenance type requirements of the current sub-cluster, each dimension corresponds to a maintenance type within the sub-cluster. The vector value is quantitatively represented based on the maintenance worker's mastery of that type of skill (e.g., whether he or she possesses basic maintenance capabilities or extensive experience), intuitively reflecting the degree of skill matching. The efficiency factor calculated based on historical work orders is a quantitative indicator derived from factors such as the ratio of the maintenance worker's actual processing time to the expected processing time, and user ratings of service efficiency. It is used to reflect the efficiency level of the maintenance worker in handling maintenance work.

[0037] A three-layer graph attention neural network is used to process the bipartite graph. The geographic location correlation, maintenance type matching, and time duration adaptability features are aggregated through message passing between nodes, and the final output is a joint embedding vector with a dimension of 128. Each embedding vector represents the comprehensive matching degree of a specific order-repairman pairing in terms of spatial accessibility, skill coverage adequacy, and work load balance. In the step, the first layer of the graph attention neural network is used to preliminarily encode the multi-dimensional feature vectors of the order node and the repairman node and extract the basic representation of each feature. The second layer of the network calculates the association weight between the order and repairman nodes through the attention mechanism (for example, the closer the geographic location, the higher the weight), and aggregates the geographic location features of both parties based on this weight to form a fused representation of the geographic location correlation. The third layer of the network focuses on the repair type matching (combining the order repair type code with the repairman skill overlap vector) and time duration adaptability (associating the order duration deviation value with the repairman efficiency factor). The two types of features are integrated through message passing, and the output features of the three layers of the network are finally fused into a 128-dimensional joint embedding vector. The dimensions of this vector correspond to quantitative representations of spatial accessibility (such as the closeness of geographical location association), skill coverage adequacy (such as the completeness of maintenance type matching), and labor load balance (such as the degree of adaptability of the maintenance worker's labor hours to handle the order), providing a comprehensive feature basis for subsequent order matching optimization.

[0038] In this embodiment, spatial clustering is used to achieve regional intensive processing of orders. A minimum number of maintenance workers is predicted based on multi-dimensional parameters such as the total number of orders, total estimated work hours, and skill diversity to lay a reasonable initial resource foundation. A bipartite graph of orders and maintenance workers is then constructed. A graph neural network is used to fuse multimodal features such as location, maintenance type, and duration to generate a joint embedding vector to quantify the overall matching degree. Finally, multiple rounds of break-and-repair cycles dynamically optimize the allocation plan, enforcing the compliance of skill and work hour constraints and dynamically expanding the maintenance worker allocation pool. This results in an intelligent order matching mechanism that balances spatial rationality, skill adaptability, and work hour balance. Density clustering is used to improve the spatial efficiency of order matching to reduce cross-regional commuting costs. A minimum number of maintenance workers is predicted based on multi-dimensional parameters to optimize resource allocation and avoid redundancy or shortage. A graph neural network is used to fuse multimodal features to improve matching accuracy. A break-and-repair cycle and dynamic expansion of the allocation pool ensure compatibility of skill and work hour constraints and solution adaptability, ultimately comprehensively improving service response efficiency, reliability, and mutual satisfaction between users and maintenance workers.

[0039] Preferably, the method executes multiple rounds of damage repair cycles based on the joint embedding vector, forcing skill matching and work hour constraints to be met in each repair phase, and dynamically expanding the repair worker allocation pool when the constraints cannot be met until convergence conditions are reached, including: Use weighted probability to remove a preset proportion of existing assignments, where pairs with skill matching below a threshold or work load exceeding the limit are removed first; Reassign maintenance workers to the removed orders based on reallocation constraints, including that the maintenance worker's skill set must include all repair types for the order and that the total duration of a single maintenance worker's assigned orders must be less than the product of the maximum daily working time and the utilization threshold. When the number of consecutive failures in the repair phase reaches the preset upper limit, based on the missing skill types and the amount of work hour gap, candidate repairmen with skill coverage of more than 90% and historical work hour utilization of less than 85% will be transferred from surrounding areas to join the repairman allocation pool; When the global matching degree improvement in three consecutive cycles is less than 0.5%, it is determined that the convergence condition has been met.

[0040] In this embodiment, a preset proportion of existing assignments is removed using weighted probability, wherein pairs with skill matching below a threshold or exceeding work load are removed first. The weight of the weighted probability is determined based on the comprehensive matching represented by the joint embedding vector, and the lower the matching degree, the higher the corresponding removal probability of the existing assignment. The skill matching degree is quantitatively evaluated by the overlap ratio between the maintenance worker's skill set and the order maintenance type set. When the ratio does not meet the set standard, it is determined that the skill matching degree is insufficient. Exceeding work load refers to the total estimated duration of all orders assigned to a maintenance worker exceeding the reasonable load range corresponding to his maximum daily working time. Such pairings are given a higher priority for removal because they may cause service delays or quality degradation. The selective removal of some assignments in the above manner aims to break the local optimal solution and create optimization space for subsequent reallocation. In the process of reallocating maintenance workers to removed orders based on redistribution constraints, the execution of redistribution constraints runs through the entire matching and screening process: for the constraint that the maintenance worker's skill set must include all maintenance types of the order, each maintenance type involved in the order must be compared one by one to ensure that the maintenance worker's skill qualifications are covered, and to eliminate insufficient service capabilities caused by skill mismatch; for the constraint that the total duration of orders assigned to a single maintenance worker is less than the product of the maximum daily working time and the utilization threshold, the estimated duration of all orders currently assigned to the maintenance worker and the estimated duration of orders to be assigned must be accumulated to ensure that the accumulated value does not exceed the product of the maximum daily working time and the preset utilization threshold to avoid working time overload affecting service efficiency; by strictly following the above constraints, it is guaranteed that the reallocated orders can be effectively processed and the maintenance worker's workload can be balanced.

[0041] When the number of consecutive failures in the repair phase reaches the preset upper limit, candidate repairmen are transferred from surrounding areas to join the repairman allocation pool based on the missing skill types and the labor hour gap. Continuous failure refers to multiple unsuccessful attempts to match the removed orders with repairmen who meet the reallocation constraints. The missing skill type is the type of repair required for the order that is not covered by the skill set of the repairmen in the current allocation pool. The labor hour gap is the difference between the total effective labor hours that can be provided by the repairmen in the current allocation pool and the total expected labor hours of the sub-cluster orders. The surrounding area refers to the service area that is geographically close to the order sub-cluster and has a lower commuting cost. The screening of candidate repairmen must meet the requirements that their skill sets cover a high proportion of the missing skill types and their historical labor hour utilization is in a low range to ensure that they have sufficient free time to take on new orders. By supplementing such repairmen, the resource gap in the current allocation pool can be effectively alleviated. In the step where convergence conditions are determined to be met when the improvement in global matching degree over three consecutive cycles is less than the set standard, the global matching degree is the average of the comprehensive matching degrees represented by the joint embedding vectors of all confirmed order-repairman pairs in the order subcluster, which comprehensively reflects the overall optimization level of spatial accessibility, skill coverage adequacy, and labor load balance; the improvement amplitude is calculated as the difference between the global matching degree of the subsequent round and the global matching degree of the previous round; when the improvement amplitude of multiple consecutive cycles is lower than the set standard, it indicates that the current allocation plan has approached the global optimal state, and further iterative optimization has limited effect on improving the matching quality. Stopping the loop at this time can avoid unnecessary consumption of computing resources while ensuring the matching effect, ensuring that the final output allocation plan is both efficient and stable.

[0042] This embodiment constructs a "break-repair" dynamic optimization loop, combines weighted probability removal of low-quality assignments to break local optima, strictly adheres to skill and work-hour constraints to ensure matching feasibility, dynamically expands the repair worker pool based on resource gaps to address local resource shortages, and uses global matching convergence to determine the balance between optimization effect and efficiency. This forms a closed-loop optimization mechanism that balances matching quality, constraint compatibility, and resource adaptability, achieving continuous iterative upgrades of the order-to-repair worker assignment plan. By weighted removal of low-matching assignments to break local optima, the global optimization potential of the matching plan is enhanced. Strict skill and work-hour constraints ensure the feasibility of the assignment plan, avoiding service issues caused by skill mismatch or work-hour overload. Dynamic expansion of the assignment pool effectively addresses local resource shortages and improves the matching success rate. The setting of convergence conditions ensures matching quality while avoiding invalid iterations. Ultimately, the matching plan achieves comprehensive improvements in accuracy, feasibility, and resource utilization efficiency, optimizing overall service effectiveness.

[0043] In summary, the method provided in this embodiment can at least achieve the following effects: This invention integrates spatial clustering, multidimensional resource demand prediction, graph neural network feature fusion, and a dynamic optimization loop to construct an intelligent order matching mechanism that balances spatial rationality, skill adaptability, and work-hour balance. Specifically, density clustering based on geographic location enables spatially intensive order processing, reducing the inefficiency of cross-regional scheduling. A multi-dimensional parameter approach, including total order volume, total work hours, and skill diversity, is used to predict the lower limit for the number of repair workers, providing a scientific basis for initial resource allocation. A graph neural network is then used to fuse multimodal features of the order and repair worker, such as their location, repair type, and duration, into a joint embedding vector to quantify the overall matching degree. Furthermore, a destroy-repair loop dynamically optimizes the allocation scheme, combining the mandatory satisfaction of skill and work-hour constraints with the dynamic expansion of the allocation pool to achieve continuous iterative optimization of the matching scheme, ultimately achieving efficient, accurate, and constraint-compatible order-worker matching. Density clustering is used to improve the spatial efficiency of order matching to reduce cross-regional commuting costs. The lower limit of the number of maintenance workers is predicted based on multi-dimensional parameters to optimize resource allocation and avoid redundancy or shortage. Graph neural networks are used to fuse multimodal features to improve matching accuracy. The compatibility of skill and working time constraints and the adaptability of the solution are ensured through the destruction-repair cycle and dynamic expansion of the allocation pool, ultimately comprehensively improving service response efficiency, reliability, and the mutual satisfaction of users and maintenance workers.

[0044] See also Figure 2 In one embodiment, an order matching system based on machine learning is further provided, the system comprising: Data acquisition module 100, for acquiring order data and maintenance worker data, wherein the order data includes geographic location, maintenance type set, and estimated maintenance duration, and the maintenance worker data includes skill set and maximum daily working hours; The order clustering module 200 is used to generate order subclusters based on the geographical location of the orders using a density clustering algorithm; Maintenance worker demand prediction module 300, used to predict the lower limit of the number of maintenance workers required for each order sub-cluster based on the total number of orders, total estimated working hours and skill diversity index in the order sub-cluster; A maintenance worker allocation pool initialization module 400 is used to create a maintenance worker allocation pool with the lower limit of the number of maintenance workers as the initial value; The graph construction and embedding module 500 is used to construct a bipartite graph of orders and repairmen within the sub-cluster, and generate a joint embedding vector integrating location, repair type, and duration through a graph neural network; The iterative optimization matching module 600 is used to execute multiple rounds of damage repair cycles based on the joint embedding vector, forcing the skill matching and work time constraints to be met in each repair phase, and dynamically expanding the repair worker allocation pool when the constraints cannot be met until the convergence conditions are reached; and output the final order allocation plan.

[0045] Preferably, the maintenance worker demand prediction module 300 is further configured to: Calculate the ratio of the total estimated working hours for orders in the sub-cluster to the average daily working hours of maintenance workers as the base quantity; The base number is weighted and modified according to the skill diversity index, where the skill diversity index is measured by the uniformity of the distribution of skill types; Divide the total number of orders by the average daily processing capacity of maintenance workers as a reference value; The maximum value between the weighted adjusted basic quantity and the quantity reference value shall be taken as the lower limit of the number of maintenance workers.

[0046] Preferably, the graph construction and embedding module 500 is further configured to: All orders in the sub-cluster and the repairmen in the repairmen allocation pool are treated as independent node sets of a bipartite graph, and fully connected edges are established between order nodes and repairmen nodes; Assign a multidimensional feature vector to each order node, including the normalized geographic location coordinates, the multi-hot encoding vector of the maintenance type set, and the deviation value between the expected maintenance time and the standard time; Each maintenance worker node is assigned a multi-dimensional feature vector, including the coordinates of the historical service area center, the overlap vector between the skill set and the maintenance type requirements of the current sub-cluster, and an efficiency factor calculated based on historical work orders; A three-layer graph attention neural network is used to process the bipartite graph. The geographical location correlation, maintenance type matching and time adaptability features are aggregated through message passing between nodes. The final output is a joint embedding vector with a dimension of 128, where each embedding vector represents the comprehensive matching degree of a specific order-repairman pairing in terms of spatial accessibility, skill coverage adequacy and work load balance.

[0047] It can be understood that the functions or modules included in the system provided in this embodiment can be used to execute the method described in the above method embodiment. Its specific implementation can refer to the description of the above method embodiment. For the sake of brevity, it will not be repeated here.

[0048] The present invention also provides an electronic device, including a processor and a memory, wherein the memory is used to store computer program code, and the computer program code includes computer instructions. When the processor executes the computer instructions, the electronic device executes a method as described in any one of the possible implementation modes.

[0049] The present invention also provides a computer-readable storage medium, in which a computer program is stored. The computer program includes program instructions. When the program instructions are executed by a processor of an electronic device, the processor executes a method as described in any one of the possible implementation methods described above.

[0050] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0051] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here. Those skilled in the art will also clearly understand that the descriptions of the various embodiments of the present invention have different focuses. For the convenience and brevity of description, the same or similar parts may not be repeated in different embodiments. Therefore, for parts not described or not described in detail in a certain embodiment, reference can be made to the descriptions of other embodiments.

Claims

1. An order matching method based on machine learning, characterized in that: The method comprises: Obtaining order data and maintenance worker data, wherein the order data includes geographic location, maintenance type set, and estimated maintenance duration, and the maintenance worker data includes skill set and maximum daily working hours; Based on the geographical location of the orders, order subclusters are generated using a density clustering algorithm; Based on the total number of orders, total estimated working hours, and skill diversity index within the order sub-cluster, the lower limit of the number of maintenance workers required for each order sub-cluster is predicted; Create a maintenance worker allocation pool with the lower limit of the number of maintenance workers as the initial value; Construct a bipartite graph of orders and repairmen within the sub-cluster, and generate a joint embedding vector integrating location, repair type, and duration through a graph neural network. Based on the joint embedding vector, multiple rounds of damage repair cycles are executed, and skill matching and work time constraints are enforced in each repair phase. When the constraints cannot be met, the repairman allocation pool is dynamically expanded until convergence conditions are reached; and the final order allocation plan is output.

2. The order matching method based on machine learning according to claim 1, characterized in that: The prediction of the lower limit of the number of maintenance workers required for each order subcluster based on the total number of orders, total estimated working hours, and skill diversity index within the order subcluster includes: Calculate the ratio of the total estimated working hours for orders in the sub-cluster to the average daily working hours of maintenance workers as the base quantity; The base number is weighted and modified according to the skill diversity index, where the skill diversity index is measured by the uniformity of the distribution of skill types; Divide the total number of orders by the average daily processing capacity of maintenance workers as a reference value; The maximum value between the weighted adjusted basic quantity and the quantity reference value shall be taken as the lower limit of the number of maintenance workers.

3. The order matching method based on machine learning according to claim 1, characterized in that: The bipartite graph of orders and repairmen within the sub-cluster is constructed, and a joint embedding vector integrating location, repair type, and duration is generated through a graph neural network, including: All orders in the sub-cluster and the repairmen in the repairmen allocation pool are treated as independent node sets of a bipartite graph, and fully connected edges are established between order nodes and repairmen nodes; Assign a multidimensional feature vector to each order node, including the normalized geographic location coordinates, the multi-hot encoding vector of the maintenance type set, and the deviation value between the expected maintenance time and the standard time; Each maintenance worker node is assigned a multi-dimensional feature vector, including the coordinates of the historical service area center, the overlap vector between the skill set and the maintenance type requirements of the current sub-cluster, and an efficiency factor calculated based on historical work orders; A three-layer graph attention neural network is used to process the bipartite graph. The geographical location correlation, maintenance type matching and time adaptability features are aggregated through message passing between nodes. The final output is a joint embedding vector with a dimension of 128, where each embedding vector represents the comprehensive matching degree of a specific order-repairman pairing in terms of spatial accessibility, skill coverage adequacy and work load balance.

4. The order matching method based on machine learning according to claim 1, characterized in that: The method executes multiple rounds of damage repair cycles based on the joint embedding vector, forcing skill matching and work hour constraints to be met in each repair phase, and dynamically expanding the repair worker allocation pool when the constraints cannot be met until convergence conditions are reached, including: Use weighted probability to remove a preset proportion of existing assignments, where pairs with skill matching below a threshold or work load exceeding the limit are removed first; Reassign maintenance workers to the removed orders based on reallocation constraints, including that the maintenance worker's skill set must include all repair types for the order and that the total duration of a single maintenance worker's assigned orders must be less than the product of the maximum daily working time and the utilization threshold. When the number of consecutive failures in the repair phase reaches the preset upper limit, based on the missing skill types and the amount of work hour gap, candidate repairmen with skill coverage of more than 90% and historical work hour utilization of less than 85% will be transferred from surrounding areas to join the repairman allocation pool; When the global matching degree improvement in three consecutive cycles is less than 0.5%, it is determined that the convergence condition has been met.

5. An order matching system based on machine learning, characterized in that: The system comprises: A data acquisition module is used to acquire order data and maintenance worker data. The order data includes geographic location, maintenance type set, and estimated maintenance duration. The maintenance worker data includes skill set and maximum daily working hours. The order clustering module is used to generate order subclusters based on the geographical location of the order through a density clustering algorithm; Maintenance worker demand prediction module, which is used to predict the lower limit of the number of maintenance workers required for each order subcluster based on the total number of orders, total estimated working hours and skill diversity index in the order subcluster; The maintenance worker allocation pool initialization module is used to create a maintenance worker allocation pool with the lower limit of the number of maintenance workers as the initial value; The graph construction and embedding module is used to construct a bipartite graph of orders and repairmen within a sub-cluster and generate a joint embedding vector that integrates location, repair type, and duration through a graph neural network. An iterative optimization matching module is configured to execute multiple rounds of damage repair cycles based on the joint embedding vector, enforce skill matching and labor time constraints in each repair phase, dynamically expand the repair worker allocation pool when the constraints cannot be met, and output a final order allocation plan.

6. The machine learning-based order matching system according to claim 5, characterized in that: The maintenance worker demand forecasting module is also used to: Calculate the ratio of the total estimated working hours for orders in the sub-cluster to the average daily working hours of maintenance workers as the base quantity; The base number is weighted and modified according to the skill diversity index, where the skill diversity index is measured by the uniformity of the distribution of skill types; Divide the total number of orders by the average daily processing capacity of maintenance workers as a reference value; The maximum value between the weighted adjusted basic quantity and the quantity reference value shall be taken as the lower limit of the number of maintenance workers.

7. The machine learning-based order matching system according to claim 5, characterized in that: The graph construction and embedding module is also used to: All orders in the sub-cluster and the repairmen in the repairmen allocation pool are treated as independent node sets of a bipartite graph, and fully connected edges are established between order nodes and repairmen nodes; Assign a multidimensional feature vector to each order node, including the normalized geographic location coordinates, the multi-hot encoding vector of the maintenance type set, and the deviation value between the expected maintenance time and the standard time; Each maintenance worker node is assigned a multi-dimensional feature vector, including the coordinates of the historical service area center, the overlap vector between the skill set and the maintenance type requirements of the current sub-cluster, and an efficiency factor calculated based on historical work orders; A three-layer graph attention neural network is used to process the bipartite graph. The geographical location correlation, maintenance type matching and time adaptability features are aggregated through message passing between nodes. The final output is a joint embedding vector with a dimension of 128, where each embedding vector represents the comprehensive matching degree of a specific order-repairman pairing in terms of spatial accessibility, skill coverage adequacy and work load balance.

8. An electronic device, characterized in that: include: A processor and a memory, the memory being used to store computer program code, the computer program code comprising computer instructions, and when the processor executes the computer instructions, the electronic device executes the machine learning-based order matching method as described in any one of claims 1 to 4.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which includes program instructions. When the program instructions are executed by a processor of an electronic device, the processor executes the machine learning-based order matching method described in any one of claims 1 to 4.

Images