Order allocation processing method and system based on decision tree algorithm

Abstract

According to the order allocation processing method and system based on the decision tree algorithm provided by the invention, the order allocation channel sequence is determined based on the decision tree algorithm, and the order allocation knowledge network is constructed, so that the association information between the service order and the order allocation channel can be effectively integrated, and a comprehensive and systematic knowledge architecture is provided for order allocation. The reference order allocation channel conforming to the service logic is screened out from the order allocation knowledge network, and the comprehensive cost parameter is calculated in combination with the predicted allocation instance data to determine the target order allocation channel, so that more scientific and accurate order allocation is realized, the order allocation rationality and efficiency are improved, the allocation cost is reduced, and the user experience is improved. And the overall service quality and the operation benefit can be improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an order allocation processing method and system based on a decision tree algorithm. Background Art

[0002] In today's era of rapid development of digitalization and informatization, the number of orders in the service industry has exploded, and the efficiency and rationality of order allocation are crucial to the operation and development of enterprises. Traditional order allocation methods often rely on manual experience or simple rule matching, which has many limitations.

[0003] On the one hand, the manual allocation of orders is inefficient and difficult to cope with the processing needs of massive orders. As the business scale continues to expand, the sharp increase in the number of orders makes manual processing difficult, and it is easy to have problems such as untimely and inaccurate allocation, which affects customer experience and the service quality of the enterprise.

[0004] On the other hand, the order allocation method based on simple rule matching lacks flexibility and intelligence. This method can usually only allocate orders according to pre-set fixed rules, and cannot fully consider the diverse characteristics and complex business logic of service orders. For example, different service orders may have different service requirements, customer preferences, time requirements, etc., and simple rule matching is difficult to comprehensively analyze these factors, resulting in order allocation results that may not conform to the actual service scenario, unable to achieve the optimal allocation of resources, and increasing the company's operating costs.

[0005] In addition, existing order allocation methods have poor adaptability when facing dynamically changing business environments and data. Market demand, customer behavior and other factors are constantly changing. Traditional methods are difficult to adjust and optimize in a timely manner based on real-time data, and cannot meet the accuracy and efficiency requirements of enterprises for order allocation in a complex and changing market environment. Summary of the invention

[0006] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides an order allocation processing method based on a decision tree algorithm, the method comprising:

[0007] Acquire a service order sequence, acquire attention tag data corresponding to the service orders in the service order sequence, and determine an order allocation channel sequence associated with the service order sequence through a decision tree algorithm based on the attention tag data;

[0008] Constructing index association information for the ath service order in the service order sequence and the basic order allocation channel of the ath service order in the order allocation channel sequence, and constructing an order allocation knowledge network based on the service order sequence, the order allocation channel sequence and the index association information; a is a positive integer;

[0009] In the order allocation knowledge network, the network internal path corresponding to the a-th service order is output as an order network path sequence, and the basic order allocation channel corresponding to the network internal path that conforms to the service logic in the order network path sequence is output as a reference order allocation channel corresponding to the a-th service order;

[0010] Obtaining predicted allocation instance data between each service order in the service order sequence and the reference order allocation channel corresponding to the a-th service order, and based on the predicted allocation instance data corresponding to the reference order allocation channel of each service order, obtaining the comprehensive cost parameter corresponding to the reference order allocation channel of each service order, and outputting the reference order allocation channel corresponding to the minimum comprehensive cost parameter as the target order allocation channel;

[0011] The a-th service order is allocated based on the target order allocation channel.

[0012] On the other hand, an embodiment of the present invention also provides an order allocation processing system based on a decision tree algorithm, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0013] Based on the above aspects, the embodiment of the present application determines the order allocation channel sequence based on the decision tree algorithm and constructs an order allocation knowledge network, which can effectively integrate the association information between service orders and order allocation channels, and provide a comprehensive and systematic knowledge architecture for order allocation. By screening out reference order allocation channels that conform to the service logic in the order allocation knowledge network, and combining the predicted allocation instance data to calculate the comprehensive cost parameters to determine the target order allocation channel, a more scientific and accurate order allocation is achieved, the rationality and efficiency of order allocation are improved, the allocation cost is reduced, and it helps to improve the overall service quality and operational benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the execution flow of the order allocation processing method based on the decision tree algorithm provided in an embodiment of the present invention.

[0015] Figure 2It is a schematic diagram of the hardware architecture of an order distribution processing system based on a decision tree algorithm provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0016] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 The figure is a flow chart of an order allocation processing method based on a decision tree algorithm provided by an embodiment of the present invention. The order allocation processing method based on a decision tree algorithm is introduced in detail below.

[0017] Step S110, obtaining a service order sequence, obtaining attention tag data corresponding to the service orders in the service order sequence, and determining an order allocation channel sequence associated with the service order sequence through a decision tree algorithm based on the attention tag data.

[0018] In this embodiment, when the financial service institution platform processes an e-commerce service order, the service order sequence must first be obtained. For example, an e-commerce platform under a large financial service institution receives a large number of service orders every day. These orders form a service order sequence according to the order of receipt or according to rules such as order type. Assume that this sequence contains multiple orders such as order A, order B, and order C.

[0019] For these service orders, each order has corresponding attention tag data. Taking order A as an example, its attention tag data may include the customer's credit rating (such as high credit rating, medium credit rating, low credit rating), the order amount range (such as small order, medium order, large order), service type (such as payment service, refund service, account query service, etc.) and the customer's membership level (such as ordinary member, silver card member, gold card member, etc.).

[0020] Based on these attention tag data, the order allocation channel sequence associated with the service order sequence is determined through the decision tree algorithm. For example, the decision tree algorithm will first divide the order according to the service type. If it is a payment service type order, the order allocation channel may be further determined based on the amount range and customer credit rating. For orders from customers with small payments and high credit ratings, they may be assigned to the fast payment channel sequence; while for orders from customers with large payments and medium credit ratings, they may be assigned to the payment channel sequence that requires manual review. For refund service type orders, if the order is for a silver card member and a small refund, it may be assigned to the automatic refund channel sequence; and for a gold card member and a large refund, it may be assigned to a special advanced customer service processing refund channel sequence. Through such layer-by-layer decision tree analysis, the order allocation channel sequence associated with the entire service order sequence is finally determined.

[0021] Step S120, constructing index association information for the ath service order in the service order sequence and the basic order allocation channel of the ath service order in the order allocation channel sequence, and constructing an order allocation knowledge network based on the service order sequence, the order allocation channel sequence and the index association information. a is a positive integer.

[0022] Assume that in the above service order sequence, the ath service order is order B. First, obtain the order attribute reference template associated with the service order sequence by the financial service institution platform. This order attribute reference template contains the reference method and format of various attribute information of the order, such as the definition and storage format of attributes such as order number, order time, and service request content.

[0023] Based on this order attribute reference template, a service order record corresponding to order B is constructed. This service order record records various attribute information of order B in detail, such as the number of order B is 12345, the order time is 10:00 on October 1, 2023, and the service request content is to pay a product fee of 500 yuan.

[0024] Then, index association information is constructed for the service order record corresponding to order B and the basic order allocation channel in the order allocation channel sequence of order B. For example, if the basic order allocation channel to which order B is assigned is the fast payment channel, the index association information may associate the order number of order B with the channel number of the fast payment channel, and also associate the order time, payment amount and other information for subsequent query and processing.

[0025] Finally, the order allocation knowledge network is generated by taking the service order records corresponding to all service orders in the service order sequence and the order allocation channel sequence as entity objects and the index association information as the internal path of the network. This order allocation knowledge network is like a huge relationship graph, in which each service order record and order allocation channel is a node, and the index association information is the path connecting these nodes, which fully describes the relationship between the service order and the order allocation channel.

[0026] Step S130, output the network internal path corresponding to the a-th service order in the order allocation knowledge network as an order network path sequence, and output the basic order allocation channel corresponding to the network internal path that conforms to the service logic in the order network path sequence as a reference order allocation channel corresponding to the a-th service order.

[0027] In the constructed order allocation knowledge network, for order B (i.e., the ath service order), the corresponding internal network path is output as an order network path sequence. This order network path sequence contains all path information from order B to its associated order allocation channel (such as the fast payment channel), including the intermediate nodes passed (such as the order verification node, payment risk assessment node, etc.) and the relationships connecting these nodes (such as the order of precedence, conditional judgment, etc.).

[0028] Get the service efficiency data corresponding to each internal network path in the order network path sequence. Take one of the internal network paths as an example. This path contains order verification nodes, payment risk assessment nodes, and fast payment channel nodes. The calculation of its service efficiency data is as follows:

[0029] Path length indicator: There are 3 connections from the order verification node to the fast payment channel node, so the path length is 3.

[0030] The number of nodes contained in the path within the network: This path contains 3 nodes, so the number of nodes is 3.

[0031] Connection strength index between nodes: The historical interaction frequency between order verification nodes and payment risk assessment nodes is relatively high, with an interaction success rate of 95% and an average interaction response time of 2 seconds; the historical interaction frequency between payment risk assessment nodes and fast payment channel nodes is also relatively high, with an interaction success rate of 98% and an average interaction response time of 1 second. The connection strength index is obtained based on a specific calculation method (such as weighted calculation).

[0032] Average processing time indicator in historical service records: Statistics are collected on the average time taken by these three nodes to process similar service orders (such as payment orders of similar amounts) in history. Assuming that the average processing time of the order verification node is 5 seconds, the average processing time of the payment risk assessment node is 3 seconds, and the average processing time of the fast payment channel node is 2 seconds, the average processing time indicator is obtained through comprehensive calculation.

[0033] Processing success rate indicator in historical service records: Calculate the success rate of these three nodes in historically processing similar service orders. Assuming that the success rate of the order verification node is 90%, the success rate of the payment risk assessment node is 92%, and the success rate of the fast payment channel node is 95%, the processing success rate indicator is obtained by comprehensive calculation.

[0034] Customer feedback data indicators: collect customer evaluation data on the services of these three nodes, such as 80% customer satisfaction for the order verification node, 85% customer satisfaction for the payment risk assessment node, and 90% customer satisfaction for the fast payment channel node. Comprehensive analysis yields customer feedback data indicators.

[0035] Based on these service efficiency data, the roaming priority corresponding to each internal path of the network is determined. For example, according to a pre-set calculation rule (such as weighted summation of various indicators), the score of each path is calculated, and the path with a high score has a high roaming priority.

[0036] The basic order allocation channel corresponding to each internal network path in the order network path sequence is compared with the service logic according to the walking priority. Assume that the service logic requires that the processing time of the payment order should be within 10 seconds and the success rate should be above 90%. If the basic order allocation channel corresponding to a certain internal network path (such as a fast payment channel) satisfies this service logic, then this fast payment channel is output as the reference order allocation channel corresponding to order B.

[0037] Step S140, obtain the predicted allocation instance data between each service order in the service order sequence and the reference order allocation channel corresponding to the a-th service order, and based on the predicted allocation instance data corresponding to the reference order allocation channel of each service order, obtain the comprehensive cost parameters corresponding to the reference order allocation channel of each service order, and output the reference order allocation channel corresponding to the minimum comprehensive cost parameter as the target order allocation channel.

[0038] In this embodiment, predicted allocation instance data between each service order in the service order sequence (such as order A, order B, order C, etc.) and the reference order allocation channel corresponding to order B (such as a fast payment channel) is obtained.

[0039] First, obtain the order node data corresponding to each service order. Taking order A as an example, its order node data includes order type as payment order, order status as pending, order priority as normal priority, order creation time as 15:00 on September 30, 2023, order expected completion time as 12:00 on October 1, 2023, and customer information associated with the order (such as customer name is Zhang San, credit rating is high credit rating, etc.).

[0040] Based on these order node data and the reference order allocation channel (fast payment channel) corresponding to order B, the instance allocation search interval is determined and the service flow data therein is obtained.

[0041] The order node data corresponding to each service order is extracted from the service order sequence to form an ordered order node data set.

[0042] For the fast payment channel, we extract its key feature information, such as the channel type is payment channel, the processing capacity is 10 orders per minute, the historical service efficiency is an average processing time of 5 seconds, the service response time is within 3 seconds, the current load status is 50% (that is, the number of orders currently being processed accounts for 50% of its processing capacity), and the supported service type is small payment, etc., and integrate these key feature information to generate a feature vector.

[0043] The instance allocation search interval is initially constructed based on the order node data set and feature vector. For example, service order node data of all payment types and small payments are filtered out according to the channel type and service type; the service efficiency threshold is set to an average processing time of 8 seconds according to the processing capacity and historical service efficiency, and service order node data with a service efficiency not less than this threshold is filtered out; the response time range is set to 25 seconds and the load condition range is set to 30%-70% according to the service response time and the current load condition, and the service order node data with a service response time within this range and a load condition within this range is finally filtered out from the service order node data filtered out for the second time, and the time period or processing node set corresponding to the finally filtered service order node data is used as the preliminary construction result of the instance allocation search interval.

[0044] Dynamically adjust the instance allocation search interval that is initially constructed. For example, based on the historical service efficiency information of the fast payment channel, if it is found that its service efficiency has been on an upward trend in the recent period, expand the scope of the instance allocation search interval according to the first preset expansion strategy; if the load corresponding to the current load status information is less than the first load (assuming that the first load is 40%), expand the scope of the instance allocation search interval according to the second preset expansion strategy; at the same time, according to the priority information of the service order in the order node data set, adjust the priority of the instance allocation search interval, such as increasing the weight of the instance allocation search interval corresponding to the high-priority order.

[0045] Determine the service flow data acquisition strategy based on the dynamically adjusted instance allocation search interval. Determine the data source information of the service flow data, such as obtaining data from the historical service record database of the financial service institution, the real-time service monitoring system, and the customer service feedback system of the e-commerce platform. According to the scope of the instance allocation search interval, set the conditions for obtaining service flow data, such as obtaining service flow data between September 25, 2023 and October 1, 2023, with the service type being payment service, the processing node being payment-related node, and the customer type being a high-credit-rating customer, and the acquisition method being batch acquisition. At the same time, determine the corresponding data encryption and desensitization strategies, as well as the storage and management methods for service flow data.

[0046] The corresponding service flow data is obtained according to the service flow data acquisition strategy, and matched with the instance allocation search interval, the service flow data that meets the instance allocation search interval range is screened out, and the service flow data that meets the instance allocation search interval range is feature extracted and analyzed to generate feature analysis results of the service flow data.

[0047] Based on the feature analysis results of the service flow data, the instances in the service flow data are screened and evaluated, the screened and evaluated service flow data instance set is optimized and recommended, and an optimized service flow data instance set is generated.

[0048] Assume that the allocation instance restriction requirements include service channel type information (such as payment channel requirement) and service response time limit information (such as response within 5 seconds). Output the order node data corresponding to each service order in the service order sequence as the initial order node, output the reference order allocation channel (fast payment channel) corresponding to order B as the termination order node, perform instance retrieval in the service flow data, and generate a basic instance sequence. If a basic instance K in the basic instance sequence does not meet the service channel type information (such as not a payment channel) or does not meet the service response time limit information (such as the response time exceeds 5 seconds), delete this basic instance K from the basic instance sequence, and output the remaining basic instances in the basic instance sequence as predicted allocation instance data.

[0049] Based on the predicted allocation instance data corresponding to the reference order allocation channel (fast payment channel) of each service order, the comprehensive cost parameters corresponding to the reference order allocation channel of each service order are obtained. The calculation of the comprehensive cost parameters may involve multiple factors, such as processing cost, risk cost, customer satisfaction impact, etc. For example, for a certain order, there may be a certain handling fee (processing cost) for processing through the fast payment channel, and there is also a very small payment risk (risk cost), but the customer satisfaction is high (good impact on customer satisfaction). The comprehensive cost parameters are calculated by combining these factors. The reference order allocation channel corresponding to the smallest comprehensive cost parameter is output as the target order allocation channel.

[0050] Step S150: Allocate the a-th service order based on the target order allocation channel.

[0051] In this embodiment, order B is allocated based on the target order allocation channel (such as the fast payment channel) determined above. The financial service institution platform directs the payment request of order B to the fast payment channel, and the payment system and risk control system related to the fast payment channel process order B according to their respective processes, such as payment verification, risk assessment, fund transfer, etc., to ensure that the service of order B can be completed efficiently, accurately and safely.

[0052] Based on the above steps, the embodiment of the present application determines the order allocation channel sequence based on the decision tree algorithm and constructs an order allocation knowledge network, which can effectively integrate the association information between service orders and order allocation channels, and provide a comprehensive and systematic knowledge architecture for order allocation. By screening out reference order allocation channels that conform to the service logic in the order allocation knowledge network, and combining the predicted allocation instance data to calculate the comprehensive cost parameters to determine the target order allocation channel, a more scientific and accurate order allocation is achieved, the rationality and efficiency of order allocation are improved, the allocation cost is reduced, and it helps to improve the overall service quality and operational benefits.

[0053] In a possible implementation, step S110 includes:

[0054] Step S111, based on the channel-related knowledge data in the attention tag data, determine the order allocation search service domain, and call the business logic tool to obtain the service channel data in the order allocation search service domain.

[0055] Step S112, obtaining the service tag knowledge data in the attention tag data, determining the service channel matching the service tag knowledge data in the service channel data in the order allocation search service domain, and loading the service channel into the order allocation channel sequence associated with the service order sequence.

[0056] In this embodiment, in the e-commerce service operation of the financial service institution, each service order carries a wealth of attention tag data. The channel-related knowledge data in these attention tag data is of great significance. For example, the channel-related knowledge data may contain potential relationship rules between different types of service orders and specific service channels. Based on these channel-related knowledge data, the order allocation search service domain can be determined. Taking the payment order of the e-commerce platform as an example, if the channel-related knowledge data indicates that high-value orders are associated with payment channels of a specific security level, then an order allocation search service domain with high-security payment services as the core can be determined. Then, the business logic tool is called to obtain the service channel data within this order allocation search service domain. The business logic tool is a tool developed by the financial service institution according to business rules, which can filter out the service channel data that meets the requirements from the huge service channel database. For example, this tool may filter out the relevant channel data in the high-security payment service domain from a large number of payment channels, refund channels and other service channels according to the set security level, service type and other conditions.

[0057] Next, obtain the service tag knowledge data in the attention tag data. Service tag knowledge data is information that describes the specific service demand characteristics of the service order. For example, for an e-commerce service order, the service tag knowledge data may include whether the service is a one-time payment, installment payment or refund operation, whether it is a domestic payment or a cross-border payment, etc. In the service channel data in the order allocation search service domain that has been obtained, determine the service channel that matches these service tag knowledge data. If the service tag knowledge data indicates that it is a small domestic one-time payment, then find a service channel suitable for small domestic one-time payment in the previously screened service channel data, such as some quick payment channels that cooperate with domestic banks. Finally, load the found service channel into the order allocation channel sequence associated with the service order sequence. This order allocation channel sequence is an ordered set that contains the possible allocation channels for each order determined according to different service order characteristics, laying the foundation for the subsequent order allocation process.

[0058] In a possible implementation, step S120 includes:

[0059] Step S121: acquiring an order attribute reference template associated with the service order sequence, and constructing a service order record corresponding to the service order in the service order sequence based on the order attribute reference template.

[0060] Step S122: construct index association information for the service order record corresponding to the ath service order in the service order sequence and the basic order allocation channel of the ath service order in the order allocation channel sequence.

[0061] Step S123, using the service order records corresponding to the service orders in the service order sequence and the order allocation channel sequence as entity objects, and using the index association information as the internal path of the network, to generate the order allocation knowledge network.

[0062] When a financial service institution processes an e-commerce service order, it must first obtain the order attribute reference template associated with the service order sequence. This order attribute reference template is a standardized definition and reference method for various attribute information of a service order. For example, it specifies the storage format and reference rules of attributes such as order number, order time, service type, and customer identity in the database. Based on this order attribute reference template, a service order record corresponding to the service order in the service order sequence is constructed. Taking a specific e-commerce service order as an example, assuming that the order number is 56789, the order time is 14:00 on November 10, 2023, the service type is cross-border payment, and the customer identity is customer number 123456. According to the order attribute reference template, this information will be constructed into a complete service order record in the prescribed format, which records the key attributes of the service order in detail.

[0063] Then, index association information is constructed for the service order record corresponding to the ath service order in the service order sequence and the basic order allocation channel of the ath service order in the order allocation channel sequence. Assume that the ath service order is the cross-border payment order with order number 56789 mentioned above, and its basic order allocation channel in the order allocation channel sequence is a certain international payment channel. The index association information will associate the key information in this service order record, such as order number, customer identity, etc., with the relevant identification information of the international payment channel. This association may include establishing an associated key-value pair in the database, or establishing a specific pointing relationship in the data structure, so that relevant information can be quickly queried and obtained.

[0064] Finally, the order allocation knowledge network is generated by taking the service order records and order allocation channel sequences corresponding to the service orders in the service order sequence as entity objects and the index association information as the internal path of the network. In the e-commerce service order processing system of this financial service institution, all service order records and order allocation channel sequences are entity objects in the network. The service order records contain detailed information of each service order, and the order allocation channel sequence contains possible allocation channels for different orders. The index association information is like a bridge connecting these entity objects, forming the internal path of the network. For example, the customer identity in a service order record is connected to the specific identifier of an order allocation channel through the index association information, forming a complete order allocation knowledge network. This order allocation knowledge network can fully reflect the complex relationship between service orders and order allocation channels, and provide a solid data foundation for subsequent order allocation decisions, path finding, and efficiency optimization operations. In this order allocation knowledge network, according to the various attribute information of the service order, the appropriate order allocation channel can be quickly located through the index association information, and the relevant service order records can also be traced back according to the relevant information of the order allocation channel, thereby realizing the efficient integration and utilization of information.

[0065] In a possible implementation, step S130 includes:

[0066] Step S131, obtaining service efficiency data corresponding to each internal network path in the order network path sequence, determining the roaming priority corresponding to each internal network path in the order network path sequence based on the service efficiency data, and comparing the basic order allocation channel corresponding to each internal network path in the order network path sequence with the service logic according to the roaming priority.

[0067] Step S132: If there is a basic order allocation channel corresponding to the internal network path in the order network path sequence that conforms to the service logic, the basic order allocation channel that conforms to the service logic is output as the reference order allocation channel corresponding to the ath service order.

[0068] In a possible implementation, step S131 includes:

[0069] Step S1311, construct a service efficiency evaluation index system, the input of which is the initial description data of each internal network path in the order network path sequence, the initial description data including the path length of the internal network path, the number of nodes contained in the internal network path, the connection strength between nodes, the average processing time in the historical service records, the processing success rate in the historical service records and customer feedback data. The output of the service efficiency evaluation index system is a set of service efficiency evaluation indicators customized for each internal network path. The service efficiency evaluation indicators in the service efficiency evaluation index set are used to measure the efficiency performance of the internal network path when processing service orders. Among them, the path length indicator is obtained by calculating the total number of steps or the total number of connections from the starting node to the ending node in the internal network path. The number of nodes contained in the internal network path directly counts the number of all nodes in the internal network path. The connection strength indicator between nodes is calculated based on the historical interaction frequency, interaction success rate and interaction response time between nodes. The average processing time indicator in the historical service records is obtained by counting the average time of all nodes in the internal network path to process the same type of service orders. The processing success rate indicator in the historical service records is obtained by calculating the success rate of all nodes in the internal network path to process the same type of service orders. The customer feedback data indicator is obtained by collecting and analyzing customer evaluation data on the services of each node in the internal network path.

[0070] Step S1312: preprocess the service efficiency evaluation indicator set and the corresponding original service data of the internal network path to generate a preprocessed service efficiency data set.

[0071] Step S1313, applying a multi-source information fusion algorithm to fuse the pre-processed service efficiency data set and the supplementary data from different sources to generate fused service efficiency data.

[0072] Step S1314, based on the fused service efficiency data, the service efficiency score of each internal network path is calculated. The specific calculation process includes setting a corresponding weight for each service efficiency evaluation indicator according to the service efficiency evaluation indicator in the service efficiency evaluation indicator set, wherein the size of the weight reflects the importance of the service efficiency evaluation indicator in evaluating service efficiency, and then using a linear weighted sum, a nonlinear model or a machine learning algorithm to combine the service efficiency data in the fused service efficiency data set with the weight of the service efficiency evaluation indicator to calculate the service efficiency score of each internal network path, and normalize the calculated service efficiency score.

[0073] Step S1315, dynamically adjust the service efficiency scores of each internal network path and the historical service efficiency data to generate a dynamically adjusted service efficiency score. Specifically, first analyze the historical service efficiency data to identify the changing trend of service efficiency over time, and set corresponding adjustment rules based on the changing trend. Then, apply the adjustment rules to dynamically adjust the service efficiency scores of each internal network path to obtain the adjusted service efficiency scores.

[0074] Step S1316, setting a calculation rule for the roaming priority according to a predefined business demand strategy, and applying the calculation rule to process the dynamically adjusted service efficiency score to calculate the roaming priority of each internal path of the network.

[0075] In this embodiment, in the e-commerce service order processing flow of the financial service institution, when it comes to the order network path sequence, the service efficiency data corresponding to each network internal path in the order network path sequence must first be obtained. Taking the processing of e-commerce payment orders by financial service institutions as an example, the order network path sequence includes each intermediate node from the order initiation node to the final payment completion and the paths connecting them. For each of the network internal paths, it is crucial to obtain the corresponding service efficiency data.

[0076] In terms of constructing a service efficiency evaluation index system, the system uses the initial description data of each internal network path in the order network path sequence as input. For example, in a typical e-commerce payment order network path, the path length of the internal network path is a key factor. The path length index is obtained by calculating the total number of steps or total number of connections from the starting node (such as the order submission node) to the ending node (such as the payment success node) in the internal network path. Assuming that an internal network path starts from order submission, passes through risk assessment, payment channel selection, identity verification and other nodes and finally reaches payment success, and the number of connections between these nodes is 5, then the path length of this internal network path is 5.

[0077] The number of nodes included in the internal path of the network directly counts the number of all nodes in the internal path of the network. Taking the internal path of the payment order above as an example, it includes five nodes: order submission, risk assessment, payment channel selection, identity verification, and payment success, so the number of nodes is 5.

[0078] The connection strength index between nodes is calculated based on the historical interaction frequency, interaction success rate and interaction response time between nodes. For example, the risk assessment node and the payment channel selection node have historically interacted 10 times per hour (historical interaction frequency), the interaction success rate is 90%, and the average interaction response time is 2 seconds. According to a specific calculation formula (comprehensively considering the weight relationship of these three factors), the connection strength index between the two nodes can be obtained.

[0079] The average processing time indicator in the historical service records is obtained by calculating the average time for all nodes in the internal path of the network to process similar service orders (such as orders with the same amount range and the same payment type). For example, the average time for the identity authentication node to process similar payment orders is 3 seconds, and the payment channel selection node is 2 seconds, etc., and the average processing time indicator of the entire internal path of the network is calculated comprehensively.

[0080] The processing success rate indicator in the historical service records is obtained by calculating the success rate of all nodes in the internal path of the network in the past processing similar service orders. Assuming that the success rate of risk assessment nodes in processing similar orders is 92%, and the payment channel selection node is 95%, etc., the processing success rate indicator of the entire internal path of the network is calculated by combining the success rates of these nodes.

[0081] Customer feedback data indicators are obtained by collecting and analyzing customer evaluation data on the services of each node in the internal path of the network. For example, customers evaluate the service satisfaction of the identity authentication node as 80%, and the service satisfaction of the payment channel selection node as 85%. The customer feedback data indicators are obtained by combining these evaluation data.

[0082] Based on these initial description data, the service efficiency evaluation index system outputs a set of service efficiency evaluation indicators customized for each internal network path. These evaluation indicators are used to measure the efficiency performance of the internal network path when processing service orders.

[0083] Next, the service efficiency evaluation index set and the original service data of the corresponding internal network path are preprocessed to generate a preprocessed service efficiency data set. The preprocessing process may include data cleaning (removing outliers, erroneous data, etc.), data standardization (converting data of different magnitudes to the same magnitude), and other operations. For example, if the processing time data of some nodes has abnormally high values ​​(possibly due to system failure recording errors), they will be identified and corrected or removed during the preprocessing process, and then the processed service efficiency evaluation index data and the original service data are integrated to generate a preprocessed service efficiency data set.

[0084] The multi-source information fusion algorithm is applied to fuse the pre-processed service efficiency data set and the supplementary data from different sources to generate the fused service efficiency data. The supplementary data may come from the evaluation data of financial service efficiency by external market research institutions, the industry average service efficiency data, etc. The multi-source information fusion algorithm will perform weighted fusion on the data from different sources based on factors such as data reliability and relevance. For example, a higher weight is given to internal historical data, and a relatively lower weight is given to external market research data, so as to generate the fused service efficiency data.

[0085] The service efficiency score of each internal network path is calculated based on the fused service efficiency data. In the specific calculation process, a corresponding weight is set for each service efficiency evaluation indicator according to the service efficiency evaluation indicator in the service efficiency evaluation indicator set. The setting of the weight reflects the importance of the service efficiency evaluation indicator in evaluating service efficiency. For example, in the payment order processing, the processing success rate indicator may be given a higher weight because it is directly related to whether the order can be successfully completed. Then, the service efficiency data in the fused service efficiency data set is combined with the weight of the service efficiency evaluation indicator using a linear weighted sum, a nonlinear model or a machine learning algorithm to calculate the service efficiency score of each internal network path. For example, using the linear weighted sum method, if the weight of the path length indicator is 0.1, the weight of the node number indicator is 0.1, the weight of the connection strength indicator is 0.2, the weight of the average processing time indicator is 0.3, the weight of the processing success rate indicator is 0.2, and the weight of the customer feedback data indicator is 0.1, the service efficiency score is calculated based on the data value and weight of each indicator. After calculating the service efficiency score, it is normalized so that the service efficiency scores of all internal network paths are within the same magnitude range for easy comparison.

[0086] Based on the service efficiency scores of each internal network path and historical service efficiency data, dynamic adjustments are made to generate dynamically adjusted service efficiency scores. First, historical service efficiency data is analyzed to identify the changing trend of service efficiency over time. For example, for a certain internal network path, it is found that its average processing time has gradually shortened over the past period of time, indicating that the service efficiency is on an upward trend. Set corresponding adjustment rules based on this changing trend. If the service efficiency increases, the weight of its service efficiency score may be appropriately increased or a certain score may be directly added to the original score; if the service efficiency decreases, the weight or score will be reduced accordingly. Apply this adjustment rule to dynamically adjust the service efficiency scores of each internal network path to obtain the adjusted service efficiency score.

[0087] The calculation rules for the roaming priority are set according to the pre-defined business demand strategy, and the dynamically adjusted service efficiency scores are processed by this calculation rule to calculate the roaming priority of each internal network path. The business demand strategy may be formulated based on the business objectives of the financial service institution, market competition, and other factors. For example, if the current focus of the financial service institution is to improve customer satisfaction, then when calculating the roaming priority, the service efficiency score corresponding to the customer feedback data indicator will be given a higher weight in the calculation rule. Through this calculation rule, the dynamically adjusted service efficiency score is calculated to obtain the roaming priority of each internal network path.

[0088] Finally, the basic order allocation channel corresponding to each network internal path in the order network path sequence is compared with the service logic according to the calculated wandering priority. The service logic is determined based on business rules, compliance requirements, customer needs and other factors in the e-commerce service order processing of financial service institutions. For example, the service logic requires that the processing time of the payment order cannot exceed 10 seconds, and the processing success rate must reach more than 90%. The network internal paths are checked from high to low according to the wandering priority. If the basic order allocation channel (such as a specific payment channel) corresponding to a certain network internal path meets this service logic, then the basic order allocation channel that meets the service logic is output as the reference order allocation channel corresponding to the a-th service order. For example, if the payment channel corresponding to a certain network internal path can complete the payment processing within 8 seconds, and the processing success rate reaches 92%, which meets the service logic requirements, then this payment channel is determined as the reference order allocation channel corresponding to the a-th service order, providing an important reference for subsequent order allocation.

[0089] In a possible implementation, step S140 includes:

[0090] Step S141, obtain the order node data corresponding to each service order in the service order sequence, determine the instance allocation search interval based on the order node data corresponding to each service order in the service order sequence and the reference order allocation channel corresponding to the a-th service order, and obtain the service flow data in the instance allocation search interval.

[0091] Step S142, based on the service flow data and the allocation instance restriction requirements in the attention tag data, perform allocation instance prediction on each service order in the service order sequence and the reference order allocation channel corresponding to the a-th service order to generate the predicted allocation instance data.

[0092] In a possible implementation, step S141 includes:

[0093] Step S1411, extracting order node data corresponding to each service order from the service order sequence, the order node data of each service order including order type, order status, order priority, order creation time, order estimated completion time and customer information associated with the order.

[0094] Step S1412, arranging the order node data in the order of the service order sequence to generate an ordered order node data set.

[0095] Step S1413, for the reference order allocation channel corresponding to the a-th service order, extract the corresponding key feature information, integrate the key feature information, and generate a feature vector, wherein the key feature information includes the channel type information, processing capability information, historical service efficiency information, service response time information, current load status information, and supported service type information of the reference order allocation channel.

[0096] Step S1414, based on the order node data set and the feature vector, a corresponding instance allocation search interval is preliminarily constructed. Specifically, according to the channel type information and service type information of the reference order allocation channel, all types of matching service order node data are screened out from the order node data set, and according to the processing capacity information and historical service efficiency information of the reference order allocation channel, a service efficiency threshold is set, and service order node data with a service efficiency not less than the service efficiency threshold is screened out again from the screened service order node data, and then according to the service response time information and current load status information of the reference order allocation channel, a response time range and a load status range are set, and finally service order node data with a service response time within the response time range and a load status within the load status range are screened out from the service order node data screened out twice, and the time period or processing node set corresponding to the finally screened service order node data is used as the preliminary construction result of the instance allocation search interval.

[0097] Step S1415, dynamically adjust the instance allocation search interval that is initially constructed to generate an instance allocation search interval that has been dynamically adjusted. Specifically, based on the historical service efficiency information in the feature vector of the reference order allocation channel, analyze the changing trend of the service efficiency over time. If the service efficiency shows an upward trend, then the range of the instance allocation search interval is expanded based on the first preset expansion strategy, otherwise the range of the instance allocation search interval is reduced based on the first preset reduction strategy. At the same time, based on the current load status information of the reference order allocation channel, if the load corresponding to the current load status information is less than the first load, then the second preset expansion strategy expands the range of the instance allocation search interval, otherwise the range of the instance allocation search interval is reduced based on the second preset reduction strategy. In addition, based on the priority information of the service order in the order node data set, the instance allocation search interval is prioritized.

[0098] Step S1416, based on the dynamically adjusted instance allocation search interval, determine the service flow data acquisition strategy. Specifically, determine the data source information of the service flow data, the source information includes the historical service record database, the real-time service monitoring system, the customer service feedback system and the third-party service data provider, and then set the service flow data acquisition conditions according to the scope of the instance allocation search interval, the acquisition conditions include the time range, service type, processing node, customer type, and then determine the service flow data acquisition method, the acquisition method includes batch acquisition, real-time acquisition, scheduled acquisition or on-demand acquisition, and determine the corresponding data encryption and desensitization strategy, as well as the storage and management method of the service flow data.

[0099] Step S1417, obtain the corresponding service flow data according to the service flow data acquisition strategy, and match the service flow data with the instance allocation search interval, filter out the service flow data that meets the instance allocation search interval range, perform feature extraction and analysis on the service flow data that meets the instance allocation search interval range, and generate feature analysis results of the service flow data.

[0100] Step S1418, based on the feature analysis results of the service flow data, the instances in the service flow data are screened and evaluated, the screened and evaluated service flow data instance set is optimized and recommended, and an optimized service flow data instance set is generated.

[0101] In a possible implementation manner, the allocation instance restriction requirement includes service channel type information and service response time limit information.

[0102] Step S142 includes:

[0103] Step S1421, output the order node data corresponding to each service order in the service order sequence as the initial order node, output the reference order allocation channel corresponding to the a-th service order as the termination order node, perform instance retrieval in the service flow data, and generate a basic instance sequence.

[0104] Step S1422: If the basic instance K in the basic instance sequence does not conform to the service channel type information, or the basic instance K does not conform to the service response time limit information, the basic instance K is deleted from the basic instance sequence, and the remaining basic instances in the basic instance sequence are output as the predicted allocation instance data.

[0105] In this embodiment, in the system of processing e-commerce service orders by financial service institutions, each service order has rich order node data. Taking the e-commerce payment order of the financial service institution as an example, the order type may be online payment, installment payment or refund. The order status may include different states such as submitted, in process, completed or failed. The order priority can be determined according to the customer's membership level, order amount or business urgency. For example, high-value orders or orders of gold card members may have higher priority. The order creation time accurately records the moment when the order is generated, which is very important for analyzing the timeliness of the order and the time management of the processing flow. The estimated completion time of the order is the estimated order completion time based on business rules and historical data. For example, for ordinary small payment orders, it is expected to be completed within 10 minutes after submission. The customer information associated with the order covers many aspects such as the customer's identity, credit level, membership level, etc. This information helps financial service institutions to classify and manage customers and provide personalized services. After extracting these order node data one by one from the service order sequence, they are arranged in the order of the service order sequence to generate an ordered order node data set.

[0106] For the reference order allocation channel corresponding to the a-th service order, the corresponding key feature information needs to be extracted and integrated into a feature vector. Assume that the a-th service order is a specific e-commerce payment order, and its corresponding reference order allocation channel is a specific payment channel. The channel type information of this payment channel indicates that it is an online payment channel, which may be a quick payment method in cooperation with a specific bank. The processing capacity information reflects the number of orders that this payment channel can process per hour, for example, 500 payment orders can be processed per hour. The historical service efficiency information contains the average processing time and success rate of similar orders processed by this payment channel in the past. For example, the average processing time for similar payment orders in the past was 5 seconds, and the success rate was 95%. The service response time information specifies the time requirement from receiving a payment request to giving a response, such as requiring a response within 3 seconds. The current load status information shows the proportion of the number of orders currently being processed by this payment channel to its processing capacity, assuming that the current load is 40%. The supported service type information indicates that this payment channel mainly supports services such as small payments and domestic payments. By integrating these key feature information, a feature vector is generated.

[0107] Based on the above order node data set and feature vector, the corresponding instance allocation search interval is preliminarily constructed. For this specific payment channel, first, according to its channel type information and service type information, all types of matching service order node data are filtered out from the order node data set. For example, if the payment channel is domestic quick payment and supports small-amount payment, then all domestic order node data related to small-amount payment orders are filtered out from the order node data set. Next, according to the processing capacity information and historical service efficiency information of the reference order allocation channel, a service efficiency threshold is set. For example, according to the historical service efficiency of this payment channel, the service efficiency threshold is set to an average processing time of no more than 8 seconds, and the service order node data with a service efficiency not less than this threshold is filtered out again from the previously filtered service order node data. Then, according to the service response time information and current load status information of the reference order allocation channel, a response time range and a load status range are set. Assuming that the service response time range is set to 2-5 seconds and the load status range is set to 30%-70%, the service order node data with a service response time within this range and a load status within this range are finally filtered out from the service order node data filtered out twice. The time period or processing node set corresponding to the finally screened service order node data is used as the preliminary construction result of the instance allocation search interval.

[0108] Subsequently, the instance allocation search interval initially constructed is dynamically adjusted. Based on the historical service efficiency information in the feature vector of the reference order allocation channel, the trend of service efficiency changes over time is analyzed. If it is found that the historical service efficiency of this payment channel shows an upward trend, for example, its average processing time has gradually decreased from 6 seconds to 5 seconds in the past few months, the range of the instance allocation search interval is expanded based on the first preset expansion strategy. This may mean that more order node data can be included for subsequent analysis, because the service efficiency of this payment channel is improving and it is capable of processing more types of orders. On the contrary, if the historical service efficiency shows a downward trend, for example, the average processing time has increased from 5 seconds to 6 seconds, the range of the instance allocation search interval is reduced based on the first preset reduction strategy to ensure that the orders processed subsequently are more in line with the actual processing capacity of this payment channel. At the same time, based on the current load status information of the reference order allocation channel, if the load corresponding to the current load status information is less than the first load (assuming that the first load is 40%), this indicates that this payment channel still has a lot of processing capacity remaining, then the range of the instance allocation search interval is expanded based on the second preset expansion strategy. If the current load is greater than the first load, for example, the current load is 60%, the range of the instance allocation search interval is reduced based on the second preset reduction strategy to avoid allocating too many orders to the payment channel that is already close to full load. In addition, the priority of the instance allocation search interval is adjusted according to the priority information of the service order in the order node data set. For high-priority orders, such as high-value orders or orders for gold card members, their weight or range in the instance allocation search interval can be appropriately expanded to ensure that these important orders can be processed first.

[0109] Based on the dynamically adjusted instance allocation search interval, determine the service flow data acquisition strategy. Determine the data source information of the service flow data, which includes the historical service record database, the real-time service monitoring system, the customer service feedback system, and the third-party service data provider. The historical service record database stores detailed processing records of all past service orders, including processing time, processing results, and service channels involved. The real-time service monitoring system can obtain the status information of the service orders currently being processed in real time, such as which orders are currently being processed on a specific payment channel and which link they have reached. The customer service feedback system collects customer evaluations and complaints about the service, which helps to understand the quality of the service and customer satisfaction. The third-party service data provider may provide some market data, industry average data, and other supplementary information. Then, according to the scope of the instance allocation search interval, set the acquisition conditions of the service flow data. The acquisition conditions include time range, such as obtaining service flow data related to the instance allocation search interval in the past month; service type, such as only obtaining flow data related to payment services; processing node, such as focusing on flow data involving specific payment channel processing nodes; customer type, such as flow data for gold card members or high credit rating customers. Then determine the acquisition method of service flow data, including batch acquisition, real-time acquisition, scheduled acquisition or on-demand acquisition. If a large amount of historical data needs to be acquired at one time for analysis, batch acquisition may be used; if the flow of currently processed orders needs to be monitored in a timely manner, real-time acquisition is used; if data is acquired at fixed time intervals (such as daily or weekly), scheduled acquisition is used; and for some special needs, such as temporary analysis of the flow data of a specific order, on-demand acquisition is used. At the same time, determine the corresponding data encryption and desensitization strategies to protect customer privacy and data security. For sensitive information, such as customer ID numbers, bank card numbers, etc., encryption technology is used to protect them, and desensitization is performed during the use of data, and only some key information is displayed. Determine the storage and management methods of service flow data, such as using a distributed storage system to classify and store data according to different service types, time ranges, etc., to facilitate subsequent query and analysis.

[0110] According to the service flow data acquisition strategy, the corresponding service flow data is obtained, and the service flow data is matched with the instance allocation search interval to filter out the service flow data that meets the instance allocation search interval range. For example, after obtaining the service flow data from the historical service record database according to the set time range, service type and other conditions, it is matched with the order node data, processing nodes, etc. within the instance allocation search interval. For the service flow data that meets the instance allocation search interval range, feature extraction and analysis are performed to generate feature analysis results of the service flow data. Feature extraction may include extracting information such as the processing time, processing results, and changes in service channels involved in the key links in the service flow process, analyzing the relationship between these features and the impact on service order processing, such as analyzing whether the extension of the processing time of a certain link will lead to the failure of the entire order processing or a decrease in customer satisfaction.

[0111] Based on the feature analysis results of the service flow data, the instances in the service flow data are screened and evaluated. For each instance in the service flow data, an evaluation is performed based on the feature analysis results obtained from the previous analysis. For example, if the processing time of a certain link in an instance is too long or the processing result is a failure, the instance may be marked or its evaluation score may be lowered. The set of service flow data instances that have been screened and evaluated is optimized and recommended to generate an optimized set of service flow data instances. The optimization process may include removing instances that obviously do not meet the requirements or have poor quality, sorting the remaining instances according to their evaluation scores, and recommending instances with higher evaluation scores as higher-quality service flow data instances.

[0112] For the process of "based on the service flow data and the allocation instance restriction requirements in the attention tag data, performing allocation instance prediction on the reference order allocation channel corresponding to each service order in the service order sequence and the a-th service order to generate the predicted allocation instance data":

[0113] Assume that the allocation instance restriction requirements include service channel type information and service response time limit information. Output the order node data corresponding to each service order in the service order sequence as the initial order node, output the reference order allocation channel corresponding to the a-th service order (such as the specific payment channel mentioned above) as the termination order node, perform instance retrieval in the service flow data, and generate a basic instance sequence. In this process, all possible instance paths from the initial order node to the termination order node are found from the service flow data, and these instance paths constitute the basic instance sequence.

[0114] If the basic instance K in the basic instance sequence does not conform to the service channel type information, for example, the basic instance K involves a refund channel instead of the required payment channel, or the basic instance K does not conform to the service response time limit information, for example, the service response time in the basic instance K exceeds the prescribed time limit, then delete the basic instance K from the basic instance sequence. After such screening, the remaining basic instances in the basic instance sequence are output as predicted allocation instance data. These predicted allocation instance data provide an important basis for the subsequent determination of the target order allocation channel, which helps financial service institutions select the most suitable and optimized order allocation plan from many possible allocation instances, thereby improving the efficiency, accuracy and customer satisfaction of e-commerce service order processing.

[0115] In a possible implementation, the method further includes:

[0116] Step A110: Send the predicted allocation instance data corresponding to the a-th service order to the order service system associated with the a-th service order, so that the order service system presents the predicted allocation instance data corresponding to the a-th service order.

[0117] Step A110, obtaining allocation response data returned by the order service system, wherein the allocation response data includes the candidate order allocation channel selected for the ath service order from the predicted allocation instance data corresponding to the ath service order.

[0118] Then the step S140 further includes:

[0119] Step S143, obtaining candidate order allocation channels corresponding to each service order in the service order sequence, and counting order support probabilities corresponding to the candidate order allocation channels.

[0120] Step S144: output the candidate order allocation channel corresponding to the maximum order support probability as the target order allocation channel, and assign and associate the target order allocation channel to each of the service orders.

[0121] In this embodiment, in the e-commerce service architecture of the financial service institution, the ath service order may be a typical e-commerce payment order. After the predicted allocation instance data corresponding to the e-commerce payment order is determined, it is necessary to send these predicted allocation instance data to the order service system associated with it. This order service system is a system specifically used to process the business logic related to the service order, which covers a series of functional modules from order reception, processing to final completion. The predicted allocation instance data contains possible allocation instance information related to the order that has been screened out after a series of complex processing, such as the relevant data of different payment channels processing the order under specific conditions. After these data are sent to the order service system, the order service system will present the predicted allocation instance data corresponding to the ath service order according to its internal display logic and business rules. This presentation process is to enable the relevant modules or personnel in the system (if there is a part involving manual intervention) to clearly see the various possible allocation instance situations for this order, so as to perform subsequent operations.

[0122] Next, obtain the allocation response data returned by the order service system. After receiving and presenting the predicted allocation instance data, the order service system will generate allocation response data based on its own business logic and possible external intervention (such as manual selection or automatic selection based on the system's internal algorithm). Taking the previous e-commerce payment order as an example, this allocation response data includes the candidate order allocation channel selected by the a-th service order in its corresponding predicted allocation instance data. For example, if the predicted allocation instance data contains three payment channels A, B, and C as possible allocation instances, and the order service system selects payment channel A based on its own load conditions, payment channel priority settings, or manually, then payment channel A becomes the choice of this order in the candidate order allocation channel, and this selection information is included in the allocation response data.

[0123] Then, based on the predicted allocation instance data corresponding to the reference order allocation channel of each service order, the comprehensive cost parameters corresponding to the reference order allocation channel of each service order are obtained, and the target order allocation channel is determined. For each service order in the service order sequence, the candidate order allocation channels corresponding to each service order must first be obtained. In the process of processing e-commerce service orders by financial service institutions, each service order has its own specific business requirements and conditions, which will affect the determination of candidate order allocation channels. For example, for a high-value e-commerce payment order, its candidate order allocation channel may tend to those payment channels with high security and high processing capabilities; while for a small payment order with high processing speed requirements, it may tend to choose a payment channel with fast processing speed and simple procedures. After obtaining the candidate order allocation channels of each service order, it is necessary to count the order support probabilities corresponding to these candidate order allocation channels.

[0124] The statistics of order support probability involve multiple factors. Taking payment orders as an example, for a payment channel as a candidate order distribution channel, its order support probability may be related to its historical processing success rate, current load status, compatibility with related business systems, and customer preferences. From the perspective of historical processing success rate, if a payment channel has a high success rate in processing similar payment orders in the past, for example, 98%, then its support probability for orders in this regard will be relatively high. The current load status will also affect the order support probability. If a payment channel is currently close to full load and has limited ability to process new orders, then its order support probability will decrease. Compatibility with related business systems is equally important. If a payment channel has good compatibility with other business systems of financial services institutions (such as risk assessment systems, customer information management systems, etc.), and can smoothly interact with data and work together, then its order support probability will increase. In addition, customer preferences will also affect the order support probability to a certain extent. If it is found from the customer's historical order data that the customer prefers to use a certain payment channel, then the support probability of this payment channel for the customer's order will also increase accordingly.

[0125] After counting the order support probabilities corresponding to each candidate order allocation channel, the candidate order allocation channel corresponding to the largest order support probability is output as the target order allocation channel. For example, there are multiple orders in the service order sequence, and each order has its own candidate order allocation channel. After counting the order support probabilities of these candidate order allocation channels, it is found that for a certain order, the order support probability of payment channel D is the highest. Then payment channel D is output as the target order allocation channel for this order. Finally, this target order allocation channel is associated with each service order, which means that in the subsequent order processing process, each service order will be processed according to this target order allocation channel. For example, the order will be directed to payment channel D for payment processing, and the relevant business logic (such as fund transfer, risk control, etc.) will also be carried out around the characteristics and requirements of payment channel D, so as to ensure the efficiency, accuracy and reliability of the entire e-commerce service order processing process. Through such a process, the target order allocation channel that best suits each service order can be selected from multiple candidate order allocation channels, thereby improving the overall efficiency of financial service institutions in processing e-commerce service orders.

[0126] In a possible implementation, the method further includes:

[0127] Step B110, when it is monitored that the service process corresponding to the target order allocation channel issues a service congestion notification, the current order node data corresponding to the service order included in the service order sequence is obtained.

[0128] Step B120: re-determine the target order allocation channel based on the order allocation knowledge network and the current order node data corresponding to the service orders included in the service order sequence.

[0129] In this embodiment, when the service process corresponding to the target order allocation channel is monitored to issue a business congestion notification, this situation indicates that the current service process has a resource shortage or insufficient processing capacity. Taking the e-commerce payment service of a financial service institution as an example, the target order allocation channel may be a specific payment channel. When the service process corresponding to this payment channel issues a business congestion notification, it means that the payment channel may be processing a large number of payment orders, which has approached or exceeded the limit of its processing capacity. At this time, it is necessary to obtain the current order node data corresponding to the service order contained in the service order sequence. Each service order in the service order sequence has its specific current order node data, which reflects various status information of the order at the current moment. For example, for an e-commerce payment order, its current order node data includes order type (such as regular payment, installment payment, etc.), order status (such as being processed, waiting for verification, etc.), order priority (determined according to factors such as customer level, amount, etc.), order creation time, order estimated completion time, and customer information associated with the order (such as customer identity, credit rating, etc.). By obtaining these detailed current order node data, the current situation of each service order can be fully understood, providing a basis for re-determining the target order allocation channel.

[0130] On this basis, the order allocation knowledge network is a network structure previously constructed that includes service order records, order allocation channel sequences, and index association information. This order allocation knowledge network structure comprehensively reflects the relationship between service orders and order allocation channels. For example, in this network, information such as order type and customer credit rating in service order records are connected to different order allocation channels through index association information. Based on this order allocation knowledge network, combined with the current order node data of the service order obtained previously, the applicability of each order allocation channel to the current service order can be re-evaluated. Taking an e-commerce payment order in a waiting verification state as an example, if its associated target order allocation channel (a payment channel) issues a business congestion notification, then other possible order allocation channels are searched in the order allocation knowledge network. If it is found that another payment channel, although its processing capacity is slightly weaker, has a lower current load and matches the order type (conventional payment) and customer credit rating (higher level) of the order, this payment channel can be re-determined as the target order allocation channel based on the comprehensive analysis of the association information in the order allocation knowledge network and the current order node data. This redetermination process is a process that comprehensively considers multiple factors. It requires in-depth analysis of various entity objects and internal network paths in the order allocation knowledge network, and at the same time combines the current specific circumstances of the service order to ensure that the redetermined target order allocation channel can effectively process service orders, avoid service delays and failures caused by business congestion, and improve the overall efficiency and reliability of e-commerce service order processing by financial services institutions.

[0131] Figure 2 The hardware structure of the order allocation processing system 100 based on the decision tree algorithm for implementing the above-mentioned order allocation processing method based on the decision tree algorithm provided in an embodiment of the present invention is shown as follows. Figure 2 As shown, the order distribution processing system 100 based on the decision tree algorithm may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0132] The machine-readable storage medium 120 may store data and / or instructions. In some embodiments, the machine-readable storage medium 120 may store data obtained from an external terminal. In some embodiments, the machine-readable storage medium 120 may store data and / or instructions that the order allocation processing system 100 based on the decision tree algorithm uses to execute or use to complete the exemplary method described in the present invention.

[0133] During the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the order allocation processing method based on the decision tree algorithm in the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0134] The specific implementation process of the processor 110 can refer to the various method embodiments executed by the order distribution processing system 100 based on the decision tree algorithm mentioned above. The implementation principles and technical effects are similar, and this embodiment will not be repeated here.

[0135] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer executable instructions are preset. When a processor executes the computer executable instructions, the above-mentioned order allocation processing method based on the decision tree algorithm is implemented.

[0136] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, various features are sometimes combined into one embodiment, drawing or description thereof.

Claims

1. An order allocation processing method based on a decision tree algorithm, characterized in that: The method comprises: Acquire a service order sequence, acquire attention tag data corresponding to the service orders in the service order sequence, and determine an order allocation channel sequence associated with the service order sequence through a decision tree algorithm based on the attention tag data; Constructing index association information for the ath service order in the service order sequence and the basic order allocation channel of the ath service order in the order allocation channel sequence, and constructing an order allocation knowledge network based on the service order sequence, the order allocation channel sequence and the index association information; a is a positive integer; In the order allocation knowledge network, the network internal path corresponding to the a-th service order is output as an order network path sequence, and the basic order allocation channel corresponding to the network internal path that conforms to the service logic in the order network path sequence is output as a reference order allocation channel corresponding to the a-th service order; Obtaining predicted allocation instance data between each service order in the service order sequence and the reference order allocation channel corresponding to the a-th service order, and based on the predicted allocation instance data corresponding to the reference order allocation channel of each service order, obtaining the comprehensive cost parameter corresponding to the reference order allocation channel of each service order, and outputting the reference order allocation channel corresponding to the minimum comprehensive cost parameter as the target order allocation channel; The a-th service order is allocated based on the target order allocation channel.

2. The order allocation processing method based on the decision tree algorithm according to claim 1 is characterized in that: The determining, based on the attention tag data, the order allocation channel sequence associated with the service order sequence by a decision tree algorithm includes: Based on the channel-related knowledge data in the attention tag data, determine the order allocation search service domain, and call the business logic tool to obtain the service channel data in the order allocation search service domain; Acquire the service tag knowledge data in the attention tag data, determine the service channel matching the service tag knowledge data in the service channel data in the order allocation search service domain, and load the service channel into the order allocation channel sequence associated with the service order sequence.

3. The order allocation processing method based on decision tree algorithm according to claim 1 is characterized in that: The step of constructing index association information for the ath service order in the service order sequence and the basic order allocation channel of the ath service order in the order allocation channel sequence, and constructing an order allocation knowledge network based on the service order sequence, the order allocation channel sequence and the index association information includes: Acquire an order attribute reference template associated with the service order sequence, and construct a service order record corresponding to the service order in the service order sequence based on the order attribute reference template; Constructing index association information for a service order record corresponding to the ath service order in the service order sequence and a basic order allocation channel of the ath service order in the order allocation channel sequence; The order allocation knowledge network is generated by taking the service order records corresponding to the service orders in the service order sequence and the order allocation channel sequence as entity objects and taking the index association information as the internal path of the network.

4. The order allocation processing method based on decision tree algorithm according to claim 1 is characterized in that: The step of outputting the basic order allocation channel corresponding to the network internal path that conforms to the service logic in the order network path sequence as the reference order allocation channel corresponding to the a-th service order comprises: Obtaining service efficiency data corresponding to each internal network path in the order network path sequence, determining a roaming priority corresponding to each internal network path in the order network path sequence based on the service efficiency data, and comparing a basic order allocation channel corresponding to each internal network path in the order network path sequence with the service logic according to the roaming priority; If there is a basic order allocation channel corresponding to the internal network path in the order network path sequence that conforms to the service logic, then the basic order allocation channel that conforms to the service logic is output as the reference order allocation channel corresponding to the a-th service order; The step of obtaining service efficiency data corresponding to each internal network path in the order network path sequence, and determining the roaming priority corresponding to each internal network path in the order network path sequence based on the service efficiency data includes: A service efficiency evaluation index system is constructed. The input of the service efficiency evaluation index system is the initial description data of each internal network path in the order network path sequence, and the initial description data includes the path length of the internal network path, the number of nodes contained in the internal network path, the connection strength between nodes, the average processing time in the historical service records, the processing success rate in the historical service records, and customer feedback data; the output of the service efficiency evaluation index system is a customized service efficiency evaluation index set for each internal network path. The service efficiency evaluation indicators in the service efficiency evaluation index set are used to measure the efficiency performance of the internal network path when processing service orders, among which the path length index is calculated by calculating the internal network path. The total number of steps or total number of connections from the starting node to the ending node in the path is obtained. The number of nodes included in the internal network path directly counts the number of all nodes in the internal network path. The connection strength index between nodes is calculated based on the historical interaction frequency, interaction success rate and interaction response time between nodes. The average processing time index in the historical service record is obtained by counting the average time of all nodes in the internal network path to process the same service orders in history. The processing success rate index in the historical service record is obtained by calculating the success rate of all nodes in the internal network path to process the same service orders in history. The customer feedback data index is obtained by collecting and analyzing customer evaluation data on the services of each node in the internal network path. Preprocessing the service efficiency evaluation indicator set and the corresponding original service data of the internal network path to generate a preprocessed service efficiency data set; Applying a multi-source information fusion algorithm to fuse the pre-processed service efficiency data set and the supplementary data from different sources to generate fused service efficiency data; Calculating the service efficiency score of each internal network path based on the fused service efficiency data, the specific calculation process includes, according to the service efficiency evaluation indicators in the service efficiency evaluation indicator set, setting a corresponding weight for each service efficiency evaluation indicator, the size of the weight reflects the importance of the service efficiency evaluation indicator in evaluating the service efficiency, and then using a linear weighted sum, a nonlinear model or a machine learning algorithm to combine the service efficiency data in the fused service efficiency data set with the weight of the service efficiency evaluation indicator, calculate the service efficiency score of each internal network path, and normalize the calculated service efficiency score; Dynamically adjust the service efficiency scores of each network internal path and historical service efficiency data to generate a dynamically adjusted service efficiency score. Specifically, first analyze the historical service efficiency data to identify the changing trend of service efficiency over time, and set corresponding adjustment rules according to the changing trend. Then, apply the adjustment rules to dynamically adjust the service efficiency scores of each network internal path to obtain an adjusted service efficiency score. The calculation rules of the roaming priority are set according to the predefined business demand strategy, and the calculation rules are applied to process the dynamically adjusted service efficiency scores to calculate the roaming priority of each internal path of the network.

5. The order allocation processing method based on decision tree algorithm according to claim 1 is characterized in that: The obtaining of predicted allocation instance data between each service order in the service order sequence and the reference order allocation channel corresponding to the a-th service order includes: Obtaining order node data corresponding to each service order in the service order sequence, determining an instance allocation search interval based on the order node data corresponding to each service order in the service order sequence and a reference order allocation channel corresponding to the a-th service order, and obtaining service flow data in the instance allocation search interval; Based on the service flow data and the allocation instance restriction requirements in the attention tag data, allocation instance prediction is performed on each service order in the service order sequence and the reference order allocation channel corresponding to the ath service order to generate the predicted allocation instance data.

6. The order allocation processing method based on decision tree algorithm according to claim 5 is characterized in that: The step of determining the instance allocation search interval based on the order node data corresponding to each service order in the service order sequence and the reference order allocation channel corresponding to the a-th service order, and acquiring the service flow data in the instance allocation search interval includes: Extracting order node data corresponding to each service order from the service order sequence, the order node data of each service order including order type, order status, order priority, order creation time, order estimated completion time, and customer information associated with the order; Arranging the order node data in the order of the service order sequence to generate an ordered order node data set; For the reference order allocation channel corresponding to the a-th service order, extract the corresponding key feature information, integrate the key feature information, and generate a feature vector, wherein the key feature information includes channel type information, processing capacity information, historical service efficiency information, service response time information, current load status information, and supported service type information of the reference order allocation channel; Based on the order node data set and the feature vector, a corresponding instance allocation search interval is preliminarily constructed; specifically, according to the channel type information and service type information of the reference order allocation channel, service order node data of all types that match are screened out from the order node data set, and according to the processing capacity information and historical service efficiency information of the reference order allocation channel, a service efficiency threshold is set, and service order node data whose service efficiency is not less than the service efficiency threshold is again screened out from the screened service order node data, and then according to the service response time information and current load status information of the reference order allocation channel, a response time range and a load status range are set, and service order node data whose service response time is within the response time range and whose load status is within the load status range is finally screened out from the service order node data screened out for the second time, and the time period or processing node set corresponding to the finally screened service order node data is used as the preliminary construction result of the instance allocation search interval; Dynamically adjust the instance allocation search interval that is initially constructed to generate an instance allocation search interval that has been dynamically adjusted; specifically, analyze the change trend of service efficiency over time based on the historical service efficiency information in the feature vector of the reference order allocation channel, and if the service efficiency shows an upward trend, expand the range of the instance allocation search interval based on the first preset expansion strategy, otherwise, reduce the range of the instance allocation search interval based on the first preset reduction strategy; at the same time, based on the current load status information of the reference order allocation channel, if the load corresponding to the current load status information is less than the first load, expand the range of the instance allocation search interval based on the second preset expansion strategy, otherwise, reduce the range of the instance allocation search interval based on the second preset reduction strategy; in addition, adjust the priority of the instance allocation search interval based on the priority information of the service order in the order node data set; Based on the dynamically adjusted instance allocation search interval, determine the service flow data acquisition strategy; specifically, determine the data source information of the service flow data, the source information includes the historical service record database, the real-time service monitoring system, the customer service feedback system and the third-party service data provider, and then set the service flow data acquisition conditions according to the scope of the instance allocation search interval, the acquisition conditions include the time range, service type, processing node, customer type, and then determine the service flow data acquisition method, the acquisition method includes batch acquisition, real-time acquisition, scheduled acquisition or on-demand acquisition, and determine the corresponding data encryption and desensitization strategy, as well as the storage and management method of the service flow data; Acquire corresponding service flow data according to the service flow data acquisition strategy, match the service flow data with the instance allocation search interval, filter out service flow data that matches the instance allocation search interval range, perform feature extraction and analysis on the service flow data that matches the instance allocation search interval range, and generate feature analysis results of the service flow data; Based on the characteristic analysis results of the service flow data, the instances in the service flow data are screened and evaluated, the screened and evaluated service flow data instance set is optimized and recommended, and an optimized service flow data instance set is generated.

7. The order allocation processing method based on decision tree algorithm according to claim 5 is characterized in that: The allocation instance restriction requirement includes service channel type information and service response time limit information; The step of performing allocation instance prediction on the reference order allocation channel corresponding to each service order in the service order sequence and the a-th service order based on the service flow data and the allocation instance restriction requirement in the attention tag data to generate the predicted allocation instance data includes: Outputting the order node data corresponding to each service order in the service order sequence as an initial order node, outputting the reference order allocation channel corresponding to the a-th service order as a termination order node, performing instance retrieval in the service flow data, and generating a basic instance sequence; If the basic instance K in the basic instance sequence does not conform to the service channel type information, or the basic instance K does not conform to the service response time limit information, the basic instance K is deleted from the basic instance sequence, and the remaining basic instances in the basic instance sequence are output as the predicted allocation instance data.

8. The order allocation processing method based on decision tree algorithm according to claim 1 is characterized in that: The method further comprises: Sending the predicted allocation instance data corresponding to the a-th service order to the order service system associated with the a-th service order, so that the order service system presents the predicted allocation instance data corresponding to the a-th service order; Acquire allocation response data returned by the order service system; the allocation response data includes a candidate order allocation channel selected by the a-th service order from the predicted allocation instance data corresponding to the a-th service order; Then, based on the predicted allocation instance data corresponding to the reference order allocation channel of each service order, the comprehensive cost parameter corresponding to the reference order allocation channel of each service order is obtained, and the reference order allocation channel corresponding to the minimum comprehensive cost parameter is output as the target order allocation channel, including: Obtain candidate order allocation channels corresponding to each service order in the service order sequence, and count order support probabilities corresponding to the candidate order allocation channels; The candidate order allocation channel corresponding to the maximum order support probability is output as the target order allocation channel, and the target order allocation channel is allocated and associated with each of the service orders.

9. The order allocation processing method based on decision tree algorithm according to claim 1 is characterized in that: The method further comprises: When monitoring that the service process corresponding to the target order allocation channel issues a service congestion notification, obtaining current order node data corresponding to the service order included in the service order sequence; The target order allocation channel is re-determined based on the order allocation knowledge network and current order node data corresponding to the service orders contained in the service order sequence.

10. An order distribution processing system based on a decision tree algorithm, characterized in that: The order allocation processing system based on the decision tree algorithm includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the order allocation processing method based on the decision tree algorithm described in any one of claims 1 to 9.

Images