Dispatching method and device, computer device and storage medium

Abstract

The embodiment of the application belongs to the field of artificial intelligence, and relates to a dispatching method and device, computer equipment and a storage medium. The method comprises the following steps: constructing a historical bipartite graph according to historical transaction information, wherein the historical bipartite graph comprises a salesperson node, a customer node and node features; calculating a recommended transaction probability between the salesperson node and the customer node based on the node features by using an initial prediction model based on a GraphSAGE algorithm; calculating a model loss of the recommended transaction probability according to a connection relationship between the salesperson node and the customer node in the bipartite graph and a prediction model, so as to adjust the initial prediction model and obtain a prediction model; updating the historical bipartite graph according to graph network update information to obtain a current bipartite graph; and calculating a recommended transaction probability between a to-be-predicted salesperson node and a to-be-predicted customer node in the current bipartite graph by using the prediction model, so as to dispatch a salesperson corresponding to the salesperson node. The application also relates to blockchain technology, and the historical transaction information can be stored in a blockchain. The application improves the accuracy of dispatching.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a dispatching method, apparatus, computer equipment, and storage medium. Background Technology

[0002] With the development of computer technology, more and more business platforms are using computers to dispatch sales personnel. In the financial and insurance sector, dispatching refers to a business platform assigning sales representatives to customers. These assigned representatives can provide services such as product recommendations. Dispatching directly impacts customer experience and product recommendation conversion rates, making it crucial for business platforms.

[0003] However, current dispatching technologies typically assign salespeople to customers randomly or based on their workload, with salespeople with lighter workloads receiving more tasks. These technologies fail to consider whether a salesperson is a good fit for a customer, resulting in low dispatching accuracy. Summary of the Invention

[0004] The purpose of this application is to provide a dispatching method, apparatus, computer equipment, and storage medium to solve the problem of low dispatching accuracy.

[0005] To address the aforementioned technical problems, this application provides a dispatching method, employing the following technical solution:

[0006] Obtain historical transaction information;

[0007] A historical bipartite graph is constructed based on the historical transaction information; the salesperson and customer in the historical transaction information correspond to the salesperson node and customer node in the historical bipartite graph, respectively, and the salesperson node and the customer node have node characteristics;

[0008] Based on the node characteristics, the probability of a successful recommendation between a salesperson node and a customer node is calculated using an initial prediction model; the initial prediction model is constructed based on the GraphSAGE algorithm.

[0009] The model loss is calculated based on the connection relationship between the salesperson node and the customer node in the bipartite graph and the recommendation success probability. The initial prediction model is then adjusted based on the model loss until the initial prediction model converges to obtain the prediction model.

[0010] The historical bipartite graph is updated based on the graph network update information to obtain the current bipartite graph;

[0011] The prediction model calculates the probability of a successful transaction between the salesperson node to be predicted and the customer node to be predicted in the current two-part diagram, and assigns tasks to the salesperson corresponding to the salesperson node based on the probability of a successful transaction.

[0012] To address the aforementioned technical problems, this application also provides a dispatching device, which employs the following technical solution:

[0013] The information acquisition module is used to acquire historical transaction information;

[0014] The bipartite graph construction module is used to construct a historical bipartite graph based on the historical transaction information; the salesperson and customer in the historical transaction information correspond to the salesperson node and customer node in the historical bipartite graph, respectively, and the salesperson node and the customer node have node characteristics;

[0015] The probability prediction module is used to calculate the probability of a successful recommendation between a salesperson node and a customer node based on the node characteristics using an initial prediction model; the initial prediction model is constructed based on the GraphSAGE algorithm.

[0016] The model adjustment module is used to calculate the model loss based on the connection relationship between the salesperson node and the customer node in the bipartite graph and the recommendation transaction probability, and adjust the initial prediction model based on the model loss until the initial prediction model converges to obtain the prediction model.

[0017] The bipartite graph update module is used to update the historical bipartite graph according to the graph network update information to obtain the current bipartite graph;

[0018] The transaction prediction module is used to calculate the probability of a recommended transaction between the salesperson node to be predicted and the customer node to be predicted in the current two-part graph through the prediction model, and to assign tasks to the salesperson corresponding to the salesperson node according to the recommended transaction probability.

[0019] To address the aforementioned technical problems, this application also provides a computer device that employs the following technical solution:

[0020] Obtain historical transaction information;

[0021] A historical bipartite graph is constructed based on the historical transaction information; the salesperson and customer in the historical transaction information correspond to the salesperson node and customer node in the historical bipartite graph, respectively, and the salesperson node and the customer node have node characteristics;

[0022] Based on the node characteristics, the probability of a successful recommendation between a salesperson node and a customer node is calculated using an initial prediction model; the initial prediction model is constructed based on the GraphSAGE algorithm.

[0023] The model loss is calculated based on the connection relationship between the salesperson node and the customer node in the bipartite graph and the recommendation success probability. The initial prediction model is then adjusted based on the model loss until the initial prediction model converges to obtain the prediction model.

[0024] The historical bipartite graph is updated based on the graph network update information to obtain the current bipartite graph;

[0025] The prediction model calculates the probability of a successful transaction between the salesperson node to be predicted and the customer node to be predicted in the current two-part diagram, and assigns tasks to the salesperson corresponding to the salesperson node based on the probability of a successful transaction.

[0026] To address the aforementioned technical problems, this application also provides a computer-readable storage medium, employing the technical solution described below:

[0027] Obtain historical transaction information;

[0028] A historical bipartite graph is constructed based on the historical transaction information; the salesperson and customer in the historical transaction information correspond to the salesperson node and customer node in the historical bipartite graph, respectively, and the salesperson node and the customer node have node characteristics;

[0029] Based on the node characteristics, the probability of a successful recommendation between a salesperson node and a customer node is calculated using an initial prediction model; the initial prediction model is constructed based on the GraphSAGE algorithm.

[0030] The model loss is calculated based on the connection relationship between the salesperson node and the customer node in the bipartite graph and the recommendation success probability. The initial prediction model is then adjusted based on the model loss until the initial prediction model converges to obtain the prediction model.

[0031] The historical bipartite graph is updated based on the graph network update information to obtain the current bipartite graph;

[0032] The prediction model calculates the probability of a successful transaction between the salesperson node to be predicted and the customer node to be predicted in the current two-part diagram, and assigns tasks to the salesperson corresponding to the salesperson node based on the probability of a successful transaction.

[0033] Compared with existing technologies, the embodiments of this application have the following main advantages: Historical transaction information is acquired and a historical bipartite graph is constructed. Salespeople and customers in the historical transaction information correspond to salesperson nodes and customer nodes in the historical bipartite graph, respectively, and these nodes possess node characteristics. Based on these node characteristics, the recommendation success probability between salesperson nodes and customer nodes is calculated using an initial prediction model constructed according to the GraphSAGE algorithm. The probability value represents the likelihood of a salesperson successfully recommending a product to a customer. The connection relationship between salesperson nodes and customer nodes can serve as a label, allowing for the calculation of model loss and adjustment of the initial prediction model until a predictive model is obtained. During application, the historical bipartite graph is updated based on graph network update information to obtain the current bipartite graph. The prediction model accurately calculates the recommendation success probability between the salesperson node to be predicted and the customer node to be predicted in the current bipartite graph. Based on the recommendation success probability, salespeople corresponding to the salesperson nodes are assigned tasks. This considers the conversion rate of product recommendations after assignment, improving the compatibility between salespeople and customers, thereby enhancing the accuracy of task assignment. Attached Figure Description

[0034] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is an exemplary system architecture diagram to which this application can be applied;

[0036] Figure 2 This is a flowchart of one embodiment of the dispatching method according to this application;

[0037] Figure 3 This is a schematic diagram of a structure of one embodiment of the dispatching device according to this application;

[0038] Figure 4 This is a schematic diagram of the structure of one embodiment of the computer device according to this application. Detailed Implementation

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0040] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0041] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0042] like Figure 1 As shown, system architecture 100 may include terminal devices 101, 102, and 103, a network 104, and a server 105. Network 104 serves as the medium for providing communication links between terminal devices 101, 102, and 103 and server 105. Network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0043] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0044] Terminal devices 101, 102, and 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 players (Moving Picture Experts Group Audio Layer IV), laptops, and desktop computers, etc.

[0045] Server 105 can be a server that provides various services, such as a backend server that supports the pages displayed on terminal devices 101, 102, and 103.

[0046] It should be noted that the dispatching method provided in this application embodiment is generally executed by the server, and correspondingly, the dispatching device is generally set in the server.

[0047] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0048] Continue to refer to Figure 2 A flowchart of an embodiment of the dispatching method according to this application is shown. The dispatching method includes the following steps:

[0049] Step S201: Obtain historical transaction information.

[0050] In this embodiment, the dispatching method operates on electronic devices (e.g., Figure 1 The server shown can communicate with the terminal via wired or wireless connection. It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultrawideband) connections, and other currently known or future wireless connection methods.

[0051] Among them, historical transaction information can record product recommendation behavior between salespersons and customers in the past, as well as product transaction behavior generated by customers purchasing products.

[0052] Specifically, historical transaction information is a record of past interactions between salespeople and customers. Salespeople and customers can interact, such as through chats and communications, where customers can consult with salespeople, and salespeople can recommend products to customers. After a consultation or product recommendation, the customer may make a purchase. All of this information is recorded and generates historical transaction information.

[0053] Historical transaction information can also include salesperson characteristics and customer characteristics. Salesperson characteristics are those related to the salesperson, such as their basic personal information, historical performance information, and profile. Customer characteristics are those related to the customer, including inquiry-related characteristics (e.g., the number of times a customer has inquired about a particular insurance product), claims-related characteristics (e.g., the claim status, amount, method of claim, and whether the claim occurred in a different location), basic information characteristics (e.g., the customer's gender, age, and region), affiliated institution characteristics (data on the customer's transactions with other subsidiaries within the group, such as the number of transactions), insured information characteristics (brand, manufacturing date, etc. of the insured), customer social characteristics (e.g., policyholder, insured, and social media activity), and customer intent characteristics (e.g., the customer's intent regarding discounts or product details during product inquiries).

[0054] It should be emphasized that, to further ensure the privacy and security of the aforementioned historical transaction information, this information can also be stored in a node of a blockchain.

[0055] The blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0056] Step S202: Construct a historical bipartite graph based on historical transaction information; the salesperson and customer in the historical transaction information correspond to the salesperson node and customer node in the historical bipartite graph, respectively, and the salesperson node and customer node have node characteristics.

[0057] Specifically, a bipartite graph is constructed based on historical transaction information. A bipartite graph, also known as a biphasic graph, is a model in graph theory. It consists of nodes and connecting edges. Nodes include salesperson nodes and customer nodes. Salesperson nodes correspond to salespeople in the historical transaction information, and customer nodes correspond to customers in the historical transaction information. If a customer purchases a product after being recommended by a salesperson, then there is a connecting edge between the salesperson node and the customer node. Furthermore, salesperson and customer nodes in the bipartite graph possess node characteristics. The node characteristics of salesperson nodes are derived from the salesperson's characteristics, and the node characteristics of customer nodes are derived from the customer's characteristics.

[0058] Step S203: Based on node features, calculate the probability of a successful recommendation between a salesperson node and a customer node using an initial prediction model; the initial prediction model is constructed based on the GraphSAGE algorithm.

[0059] The initial prediction model can be a prediction model that has not yet completed model training.

[0060] Specifically, the predictive model predicts the probability of a salesperson recommending a product to a customer, i.e., the likelihood of a customer purchasing the product after the salesperson recommends it. The initial predictive model / predictive model is built based on the GraphSAGE algorithm, a graph neural network algorithm. The core of the algorithm is to optimize the sampling of the entire graph down to the sampling of the neighboring nodes of the target node, using K neighboring nodes to calculate the attributes of the target node, which is an inductive learning model.

[0061] During training, to predict the probability of a successful recommendation between a salesperson node and a customer node, the customer node needs to be treated as a neighbor node to be collected. Then, the probability of a successful recommendation between the salesperson node and the customer node is calculated using an initial prediction model based on the GraphSAGE algorithm. The node features of the nodes involved will be used as input to the GraphSAGE algorithm to complete the calculation of the probability of a successful recommendation.

[0062] Step S204: Calculate the model loss based on the connection relationship between the salesperson node and the customer node in the bipartite graph and the recommendation success probability, and adjust the initial prediction model based on the model loss until the initial prediction model converges to obtain the prediction model.

[0063] Specifically, the bipartite graph contains labels required for model training. If there is a connection between a salesperson node and a customer node, it indicates that the salesperson has successfully recommended the product to the customer, and the label is 1; if there is no connection between a salesperson node and a customer node, it indicates that the salesperson has not successfully recommended the product to the customer, and the label is 0.

[0064] The loss function during training can be the cross-entropy loss function. The model loss can be calculated based on the labels and the probability of successful recommendation. With minimizing the model loss as the training objective, the model parameters of the initial prediction model are adjusted, and the initial prediction model with adjusted parameters is iteratively trained until the initial prediction model converges, at which point training stops, and the prediction model is obtained.

[0065] Step S205: Update the historical bipartite graph according to the graph network update information to obtain the current bipartite graph.

[0066] Specifically, the historical bipartite graph is constructed based on historical data. In application, new salespersons and new customers may join. At this time, it is necessary to obtain graph network update information to update the historical bipartite graph and obtain the current bipartite graph.

[0067] Step S206: Calculate the probability of a sales transaction between the salesperson node to be predicted and the customer node to be predicted in the current two-part graph using the prediction model, and assign tasks to the salespersons corresponding to the salesperson nodes based on the probability of a sales transaction.

[0068] Specifically, the nodes in the current two-part diagram are the salesperson nodes to be predicted and the customer nodes to be predicted. The prediction model can predict the probability of a successful referral between each salesperson node and each customer node to be predicted.

[0069] It's understandable that a higher recommendation-to-sell probability means a higher likelihood of a customer purchasing the product after the salesperson recommends it. Therefore, salespeople can be assigned based on this probability; that is, the higher the recommendation-to-sell probability between a salesperson and a customer, the more likely the salesperson will be assigned to that customer.

[0070] In this embodiment, historical transaction information is acquired and a historical bipartite graph is constructed. Salespeople and customers in the historical transaction information correspond to salesperson nodes and customer nodes in the historical bipartite graph, respectively, and these nodes possess node characteristics. Based on these node characteristics, the probability of a successful recommendation between a salesperson node and a customer node is calculated using an initial prediction model constructed according to the GraphSAGE algorithm. The probability value represents the likelihood of a salesperson successfully recommending a product to a customer. The connection relationship between salesperson nodes and customer nodes can serve as a label, allowing for the calculation of model loss and adjustment of the initial prediction model until a final prediction model is obtained. During application, the historical bipartite graph is updated based on graph network update information to obtain the current bipartite graph. The prediction model accurately calculates the probability of a successful recommendation between the salesperson node to be predicted and the customer node to be predicted in the current bipartite graph. Based on this probability, salespeople corresponding to the salesperson nodes are assigned tasks, taking into account the conversion rate of product recommendations after assignment, thus improving the compatibility between salespeople and customers and consequently increasing the accuracy of task assignment.

[0071] Furthermore, step S202 may include: identifying salespersons and customers in historical transaction information, as well as the referral relationship between salespersons and customers; extracting salesperson characteristics of salespersons and customer characteristics of customers from historical transaction information; designating salespersons and customers as salesperson nodes and customer nodes respectively, and connecting salesperson nodes and customer nodes with referral relationships; determining salesperson characteristics as node characteristics of the corresponding salesperson nodes, and determining customer characteristics as node characteristics of the corresponding customer nodes, to obtain a historical bipartite graph.

[0072] Specifically, the data stored in historical transaction information is tagged, which can identify the salespersons and customers recorded within, as well as the referral relationship between them. If a salesperson successfully recommends a product to a customer, a referral relationship is established between the salesperson and the customer. Furthermore, the salesperson's characteristics and the customer's characteristics can be extracted from the historical transaction information.

[0073] When constructing the historical bipartite graph, salespeople are designated as salesperson nodes, and customers as customer nodes. If a referral relationship exists between a salesperson and a customer, a connecting edge connects the salesperson node and the customer node. The connecting edge can have an identifier to indicate its type, and can be established based on the referral relationship. In one embodiment, if historical transaction information determines that a salesperson and a customer have previously engaged in chat or consultation, a connecting edge can also be established between the salesperson node and the customer node, and this type of connecting edge can be distinguished from the aforementioned connecting edges established based on the referral relationship using a connecting edge identifier.

[0074] Then, the salesperson characteristics of the salesperson corresponding to the salesperson node are determined as the node characteristics of that salesperson node, and the customer characteristics of the customer corresponding to the customer node are determined as the node characteristics of that customer node, thus completing the construction of the historical bipartite graph.

[0075] In this embodiment, the salesperson and customer in the historical transaction information are respectively designated as salesperson nodes and customer nodes. The salesperson nodes and customer nodes with a recommendation-to-transaction relationship are connected, and node features are added to the salesperson nodes and customer nodes respectively, thereby completing the construction of the historical bipartite graph and preparing for model training.

[0076] Furthermore, after step S202, the method may further include: calculating the density of the historical bipartite graph; when the density is less than a preset density threshold, adjusting the historical bipartite graph according to a preset density increase strategy.

[0077] Among them, density reflects the density of the historical bipartite graph in numerical form. The higher the density, the more information is available in the historical bipartite graph.

[0078] Specifically, there are various ways to calculate the density. In one embodiment, the total number of nodes and the number of non-isolated nodes in the historical bipartite graph are obtained, and the ratio of the number of non-isolated nodes to the total number of nodes is used as the density. Here, an isolated node can be a node in the historical bipartite graph without any connecting edges, that is, the node is not connected to any other node through any connecting edge. For a non-isolated node, the node is connected to other nodes through at least one connecting edge.

[0079] In one embodiment, based on the number of salesperson nodes and customer nodes in the historical bipartite graph, the maximum number of connection edges that can be established between salesperson nodes and customer nodes is calculated. Then, the number of existing connection edges between salesperson nodes and customer nodes in the historical bipartite graph is counted, and the ratio of the number of existing connection edges to the maximum number is used as the density.

[0080] A preset density threshold is obtained, and the calculated density is compared with the threshold. If the density is less than the threshold, it indicates that there is less usable information in the historical bipartite graph. In this case, the historical bipartite graph can be adjusted according to a preset density-increasing strategy to increase its density. Adjusting the historical bipartite graph mainly involves adjusting the connecting edges, such as increasing the number of connecting edges.

[0081] In this embodiment, the density of the historical bipartite graph is calculated. When the density is less than the density threshold, it indicates that there is less usable information in the historical bipartite graph. The historical bipartite graph is adjusted according to the preset density increase strategy to enrich the information in the historical bipartite graph and ensure the accuracy of model training.

[0082] Furthermore, the steps for adjusting the historical bipartite graph according to the preset density increase strategy may include: calculating the similarity of similar nodes in the historical bipartite graph based on node features; when there is a similarity of similar nodes greater than the similarity threshold, connecting the two nodes corresponding to the similarity of similar nodes through connecting edges; or, establishing weak connecting edges between salesperson nodes and customer nodes in the historical bipartite graph based on the interaction information in historical transaction information; the edge weight of the weak connecting edge is lower than the edge weight of the connecting edge established based on the recommendation transaction relationship; the edge weight participates in the calculation of the recommendation transaction probability through node features.

[0083] Specifically, there are two types of nodes in the historical diagram: salesperson nodes and customer nodes. Two salesperson nodes are of the same type, and two customer nodes are also of the same type.

[0084] For two nodes belonging to the same category, the similarity between them can be calculated based on their node features. The higher the similarity, the more similar the two nodes are. During the calculation, the node features of each node are arranged and transformed into feature vectors. The similarity between the two nodes is then measured by the similarity between their feature vectors.

[0085] It can calculate the similarity of nodes of the same type in any group. If the similarity of nodes of the same type is greater than the preset similarity threshold, then connect the two nodes in the group of nodes of the same type through connecting edges.

[0086] In another embodiment, interaction information from historical transaction data is obtained. This interaction information can record chat and consultation behaviors between salespersons and customers. If the interaction information determines that there was a chat or consultation between the salesperson and the customer, but no product recommendation was successfully made, a weak connection edge can be established between the salesperson node and the customer node. The establishment of a weak connection edge introduces the concept of edge weight. If the connection edge is established based on a recommendation-based transaction relationship, its edge weight is greater than that of a weak connection edge. Edge weights can be applied to the node characteristics of the nodes connected by the connection edge. During calculation, edge weights can be assigned to the node characteristics of the nodes connected by the connection edge; the lower the edge weight, the less influence the node weight of the connected node plays in the calculation.

[0087] In one embodiment, the connection edges established based on the similarity of nodes of the same type can also have edge weights, and these edge weights are lower than the edge weights of the connection edges established based on the recommendation transaction relationship.

[0088] In this embodiment, the number of connecting edges in the historical bipartite graph can be increased, or edge weights can be added to the connecting edges, thereby enriching the information contained in the historical bipartite graph.

[0089] Furthermore, step S203 may include: randomly collecting one-hop neighbor nodes and two-hop neighbor nodes of the salesperson node through the initial prediction model; generating a node embedding representation of the salesperson node based on the node features of the one-hop neighbor nodes and two-hop neighbor nodes; inputting the node embedding representation into the fully connected layer of the initial prediction model to obtain the recommendation transaction probability between the salesperson node and the customer node; the customer node is a node among the one-hop neighbor nodes or two-hop neighbor nodes.

[0090] Specifically, the initial prediction model is built based on the GraphSAGE algorithm. During prediction, for a given salesperson node, its one-hop and two-hop neighbor nodes are randomly collected. Based on the node features of these one-hop and two-hop neighbor nodes, a node embedding representation of the salesperson node is generated.

[0091] In one embodiment, the node features of two-hop neighbor nodes are first aggregated to generate the node embedding representation of one-hop neighbor nodes. Then, the node embedding representations of one-hop neighbor nodes are aggregated to calculate the node embedding representation of the salesperson node.

[0092] In one embodiment, information can be aggregated from all neighboring nodes of the salesperson node, thus without information loss.

[0093] After generating node embedding representations, these representations are input into the fully connected layer of the initial prediction model. The fully connected layer then outputs the probability of a successful transaction between the salesperson node and the customer node. The customer node must be one of the one-hop or two-hop neighbors during sampling; that is, the customer node needs to be sampled during the sampling process.

[0094] In this embodiment, the neighboring nodes of the salesperson node are collected, and the node embedding representation of the salesperson node is calculated based on the node characteristics. This representation is then input into the fully connected layer of the initial prediction model to obtain the probability of a recommended transaction between the salesperson node and the customer node. This enables the prediction of the likelihood of a product recommendation transaction between the salesperson and the customer, and also measures the matching degree between the salesperson and the customer.

[0095] Furthermore, step S205 may include: obtaining bipartite graph update information; adjusting the salesperson nodes and their corresponding node features, as well as the customer nodes and their corresponding node features in the historical bipartite graph based on the node update information and node feature update information in the bipartite graph update information, to obtain the current bipartite graph.

[0096] Specifically, the update information of the bipartite graph is obtained, including node update information and node feature update information. This allows for the adjustment of salesperson nodes and their features in the historical bipartite graph to obtain the current bipartite graph. Adjustments to salesperson nodes can include adding and deleting nodes, while adjustments to node features include adding, deleting, and modifying features.

[0097] During adjustments, multiple new salesperson nodes and customer nodes can be added at once to enable the prediction model to forecast new salespersons and new customers.

[0098] In this embodiment, the nodes and node features in the historical bipartite graph are adjusted according to the bipartite graph update information, so that the prediction model can make predictions based on the current bipartite graph.

[0099] Furthermore, step S206 may include: calculating the ranking value of each salesperson based on historical transaction information; sorting each salesperson according to the ranking value to obtain a salesperson queue; and assigning tasks to each salesperson in the salesperson queue based on the recommended transaction probability of the salesperson node corresponding to the salesperson.

[0100] Specifically, a ranking value is calculated for each salesperson based on historical transaction information. This ranking value can be the number of times a salesperson successfully recommended products within a preset time period, or the total value of the successfully recommended products. The ranking value can also be calculated based on the node characteristics of the salesperson's node.

[0101] It's understandable that a higher ranking value indicates a salesperson has successfully recommended products more times, or a higher total value of the recommended products. Salespeople are sorted in descending order based on their ranking values ​​to create a salesperson queue. During task assignment, salespeople are selected from the queue from front to back, and tasks are assigned to them based on the probability of a sale made through each recommendation at their respective salesperson node. That is, salespeople with higher ranking values ​​are assigned tasks first. During task assignment, salespeople can either choose customers based on their recommendation success rate or be assigned to customers whose success rate is higher than a preset probability value.

[0102] In this embodiment, the ranking value of the salesperson is calculated, and the salespersons are ranked and assigned tasks according to the ranking value. Salespersons with higher ranking values ​​can be assigned customers first, which improves the accuracy and success rate of product recommendations after task assignment.

[0103] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0104] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0105] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware with computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When executed, the program can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0106] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0107] Further reference Figure 3 As a response to the above Figure 2 To implement the method shown, this application provides an embodiment of a dispatching device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0108] like Figure 3 As shown, the dispatching device 300 described in this embodiment includes: an information acquisition module 301, a bipartite graph construction module 302, a probability prediction module 303, a model adjustment module 304, a bipartite graph update module 305, and a transaction prediction module 306, wherein:

[0109] The information acquisition module 301 is used to acquire historical transaction information.

[0110] The bipartite graph construction module 302 is used to construct a historical bipartite graph based on historical transaction information. The salesperson and customer in the historical transaction information correspond to the salesperson node and customer node in the historical bipartite graph, respectively, and the salesperson node and customer node have node characteristics.

[0111] The probability prediction module 303 is used to calculate the probability of a recommended transaction between a salesperson node and a customer node based on node features and through an initial prediction model; the initial prediction model is built based on the GraphSAGE algorithm.

[0112] The model adjustment module 304 is used to calculate the model loss based on the connection relationship between the salesperson node and the customer node in the bipartite graph and the recommendation success probability, and adjust the initial prediction model based on the model loss until the initial prediction model converges to obtain the prediction model.

[0113] The bipartite graph update module 305 is used to update the historical bipartite graph according to the graph network update information to obtain the current bipartite graph.

[0114] The transaction prediction module 306 is used to calculate the probability of a recommended transaction between the salesperson node to be predicted and the customer node to be predicted in the current two-part graph through the prediction model, and to assign tasks to the salesperson corresponding to the salesperson node based on the probability of a recommended transaction.

[0115] In this embodiment, historical transaction information is acquired and a historical bipartite graph is constructed. Salespeople and customers in the historical transaction information correspond to salesperson nodes and customer nodes in the historical bipartite graph, respectively, and these nodes possess node characteristics. Based on these node characteristics, the probability of a successful recommendation between a salesperson node and a customer node is calculated using an initial prediction model constructed according to the GraphSAGE algorithm. The probability value represents the likelihood of a salesperson successfully recommending a product to a customer. The connection relationship between salesperson nodes and customer nodes can serve as a label, allowing for the calculation of model loss and adjustment of the initial prediction model until a final prediction model is obtained. During application, the historical bipartite graph is updated based on graph network update information to obtain the current bipartite graph. The prediction model accurately calculates the probability of a successful recommendation between the salesperson node to be predicted and the customer node to be predicted in the current bipartite graph. Based on this probability, salespeople corresponding to the salesperson nodes are assigned tasks, taking into account the conversion rate of product recommendations after assignment, thus improving the compatibility between salespeople and customers and consequently increasing the accuracy of task assignment.

[0116] In some optional implementations of this embodiment, the bipartite graph construction module 302 may include: an information recognition submodule, a feature extraction submodule, a node determination submodule, and a feature determination submodule, wherein:

[0117] The information recognition submodule is used to identify salespersons and customers in historical transaction information, as well as the referral and transaction relationships between salespersons and customers.

[0118] The feature extraction submodule is used to extract salesperson features of salespeople and customer features of customers from historical transaction information.

[0119] The node determination submodule is used to designate salespersons and customers as salesperson nodes and customer nodes respectively, and connect salesperson nodes and customer nodes that have a referral and transaction relationship.

[0120] The feature determination submodule is used to determine the characteristics of salespersons as the node characteristics of the corresponding salesperson nodes and the characteristics of customers as the node characteristics of the corresponding customer nodes, thereby obtaining a historical bipartite graph.

[0121] In this embodiment, the salesperson and customer in the historical transaction information are respectively designated as salesperson nodes and customer nodes. The salesperson nodes and customer nodes with a recommendation-to-transaction relationship are connected, and node features are added to the salesperson nodes and customer nodes respectively, thereby completing the construction of the historical bipartite graph and preparing for model training.

[0122] In some optional implementations of this embodiment, the dispatching device 300 may further include: a density calculation module and a bipartite graph adjustment module, wherein:

[0123] The density calculation module is used to calculate the density of historical bipartite graphs.

[0124] The bipartite graph adjustment module is used to adjust the historical bipartite graph according to a preset density increase strategy when the density is less than a preset density threshold.

[0125] In this embodiment, the density of the historical bipartite graph is calculated. When the density is less than the density threshold, it indicates that there is less usable information in the historical bipartite graph. The historical bipartite graph is adjusted according to the preset density increase strategy to enrich the information in the historical bipartite graph and ensure the accuracy of model training.

[0126] In some optional implementations of this embodiment, the bipartite graph adjustment module may include: a similarity calculation submodule, a node connection submodule, or an establishment submodule, wherein:

[0127] The similarity calculation submodule is used to calculate the similarity of similar nodes in a historical bipartite graph based on node features.

[0128] The node connection submodule is used to connect two nodes with the same similarity to each other through a connection edge when there are nodes of the same type with similarity greater than the similarity threshold.

[0129] A submodule is established to create weak connections between salesperson nodes and customer nodes in the historical bipartite graph based on interaction information in historical transaction data. The edge weight of the weak connection is lower than that of the connection established based on the recommendation-based transaction relationship. The edge weight is used to calculate the recommendation-based transaction probability through node features.

[0130] In this embodiment, the number of connecting edges in the historical bipartite graph can be increased, or edge weights can be added to the connecting edges, thereby enriching the information contained in the historical bipartite graph.

[0131] In some optional implementations of this embodiment, the probability prediction module 303 may include: a node acquisition submodule, a representation generation submodule, and a probability prediction submodule, wherein:

[0132] The node acquisition submodule is used to randomly acquire the one-hop neighbor nodes and two-hop neighbor nodes of the salesperson node through the initial prediction model.

[0133] The representation generation submodule is used to generate the node embedding representation of the businessperson node based on the node features of the one-hop neighbor node and the two-hop neighbor node.

[0134] The probability prediction submodule is used to embed nodes into the fully connected layer that represents the input of the initial prediction model to obtain the probability of a recommended transaction between the salesperson node and the customer node; the customer node is a node among the one-hop neighbor nodes or two-hop neighbor nodes.

[0135] In this embodiment, the neighboring nodes of the salesperson node are collected, and the node embedding representation of the salesperson node is calculated based on the node characteristics. This representation is then input into the fully connected layer of the initial prediction model to obtain the probability of a recommended transaction between the salesperson node and the customer node. This enables the prediction of the likelihood of a product recommendation transaction between the salesperson and the customer, and also measures the matching degree between the salesperson and the customer.

[0136] In some optional implementations of this embodiment, the bipartite graph update module 305 may include: an update acquisition submodule and a node adjustment submodule, wherein:

[0137] The update acquisition submodule is used to obtain bipartite graph update information.

[0138] The node adjustment submodule is used to adjust the salesperson nodes and their corresponding node features, as well as the customer nodes and their corresponding node features in the historical bipartite graph update information, based on the node update information and node feature update information in the bipartite graph update information, to obtain the current bipartite graph.

[0139] In this embodiment, the nodes and node features in the historical bipartite graph are adjusted according to the bipartite graph update information, so that the prediction model can make predictions based on the current bipartite graph.

[0140] In some optional implementations of this embodiment, the transaction prediction module 306 may include: a sorting calculation submodule, a queue generation submodule, and a dispatch processing submodule, wherein:

[0141] The sorting calculation submodule is used to calculate the sorting value of each salesperson based on historical transaction information.

[0142] The queue generation submodule is used to sort each salesperson according to the sorting value to obtain the salesperson queue.

[0143] The task dispatching submodule is used to dispatch tasks to each salesperson in the salesperson queue based on the recommended transaction probability of the corresponding salesperson node.

[0144] In this embodiment, the ranking value of the salesperson is calculated, and the salespersons are ranked and assigned tasks according to the ranking value. Salespersons with higher ranking values ​​can be assigned customers first, which improves the accuracy and success rate of product recommendations after task assignment.

[0145] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4This is a basic structural block diagram of the computer device in this embodiment.

[0146] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected via a system bus. It should be noted that only the computer device 4 with components 41-43 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0147] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0148] The memory 41 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 41 may be an internal storage unit of the computer device 4, such as the hard disk or memory of the computer device 4. In other embodiments, the memory 41 may also be an external storage device of the computer device 4, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 4. Of course, the memory 41 may also include both the internal storage unit and its external storage device of the computer device 4. In this embodiment, the memory 41 is typically used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions for dispatching methods. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or will be output.

[0149] In some embodiments, the processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 42 is typically used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to execute computer-readable instructions stored in the memory 41 or to process data, for example, to execute computer-readable instructions of the dispatching method.

[0150] The network interface 43 may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the computer device 4 and other electronic devices.

[0151] The computer device provided in this embodiment can execute the above-described dispatching method. The dispatching method here can be any of the dispatching methods described in the various embodiments above.

[0152] In this embodiment, historical transaction information is acquired and a historical bipartite graph is constructed. Salespeople and customers in the historical transaction information correspond to salesperson nodes and customer nodes in the historical bipartite graph, respectively, and these nodes possess node characteristics. Based on these node characteristics, the probability of a successful recommendation between a salesperson node and a customer node is calculated using an initial prediction model constructed according to the GraphSAGE algorithm. The probability value represents the likelihood of a salesperson successfully recommending a product to a customer. The connection relationship between salesperson nodes and customer nodes can serve as a label, allowing for the calculation of model loss and adjustment of the initial prediction model until a final prediction model is obtained. During application, the historical bipartite graph is updated based on graph network update information to obtain the current bipartite graph. The prediction model accurately calculates the probability of a successful recommendation between the salesperson node to be predicted and the customer node to be predicted in the current bipartite graph. Based on this probability, salespeople corresponding to the salesperson nodes are assigned tasks, taking into account the conversion rate of product recommendations after assignment, thus improving the compatibility between salespeople and customers and consequently increasing the accuracy of task assignment.

[0153] This application also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the dispatching method as described above.

[0154] In this embodiment, historical transaction information is acquired and a historical bipartite graph is constructed. Salespeople and customers in the historical transaction information correspond to salesperson nodes and customer nodes in the historical bipartite graph, respectively, and these nodes possess node characteristics. Based on these node characteristics, the probability of a successful recommendation between a salesperson node and a customer node is calculated using an initial prediction model constructed according to the GraphSAGE algorithm. The probability value represents the likelihood of a salesperson successfully recommending a product to a customer. The connection relationship between salesperson nodes and customer nodes can serve as a label, allowing for the calculation of model loss and adjustment of the initial prediction model until a final prediction model is obtained. During application, the historical bipartite graph is updated based on graph network update information to obtain the current bipartite graph. The prediction model accurately calculates the probability of a successful recommendation between the salesperson node to be predicted and the customer node to be predicted in the current bipartite graph. Based on this probability, salespeople corresponding to the salesperson nodes are assigned tasks, taking into account the conversion rate of product recommendations after assignment, thus improving the compatibility between salespeople and customers and consequently increasing the accuracy of task assignment.

[0155] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0156] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

Claims

1. A method for dispatching work, characterized in that, Includes the following steps: Obtain historical transaction information; A historical bipartite graph is constructed based on the historical transaction information; the salesperson and customer in the historical transaction information correspond to the salesperson node and customer node in the historical bipartite graph, respectively, and the salesperson node and the customer node have node characteristics; Based on the node characteristics, the probability of a successful transaction between a salesperson node and a customer node is calculated using an initial prediction model. The initial prediction model is constructed based on the GraphSAGE algorithm; The model loss is calculated based on the connection relationship between the salesperson node and the customer node in the bipartite graph and the recommendation transaction probability. The initial prediction model is then adjusted based on the model loss until the initial prediction model converges to obtain the prediction model. The historical bipartite graph is updated based on the graph network update information to obtain the current bipartite graph; The prediction model calculates the probability of a sales transaction between the salesperson node to be predicted and the customer node to be predicted in the current two-part diagram, and assigns tasks to the salesperson corresponding to the salesperson node based on the probability of a sales transaction. The calculation of the recommendation-based transaction probability between salesperson nodes and customer nodes using the initial prediction model includes: The initial prediction model is used to randomly collect the one-hop neighbor nodes and two-hop neighbor nodes of the salesperson node; Based on the node features of the one-hop neighbor node and the two-hop neighbor node, a node embedding representation of the salesperson node is generated. The node is embedded into the fully connected layer of the initial prediction model to obtain the recommendation success probability between the salesperson node and the customer node; the customer node is a node among the one-hop neighbor nodes or the two-hop neighbor nodes.

2. The dispatching method according to claim 1, characterized in that, The step of constructing a historical bipartite graph based on the historical transaction information includes: Identify the salespersons and customers in the historical transaction information, as well as the referral and transaction relationships between the salespersons and the customers; Extract the salesperson's characteristics and the customer's characteristics from the historical transaction information; The salesperson and the customer are designated as salesperson nodes and customer nodes, respectively, and the salesperson nodes and customer nodes with a referral and transaction relationship are connected. The salesperson characteristics are determined as the node characteristics of the corresponding salesperson node, and the customer characteristics are determined as the node characteristics of the corresponding customer node, thus obtaining a historical bipartite graph.

3. The dispatching method according to claim 1, characterized in that, Following the step of constructing a historical bipartite graph based on the historical transaction information, the method further includes: Calculate the density of the historical bipartite graph; When the density is less than a preset density threshold, the historical bipartite graph is adjusted according to a preset density increase strategy.

4. The dispatching method according to claim 3, characterized in that, The step of adjusting the historical bipartite graph according to the preset density increase strategy includes: Based on node features, the similarity of similar nodes in the historical bipartite graph is calculated; When there are nodes of the same type with a similarity greater than the similarity threshold, connect the two nodes corresponding to the similarity threshold through a connecting edge; or, Based on the interaction information in the historical transaction information, weak connection edges are established between the salesperson nodes and customer nodes in the historical bipartite graph; the edge weight of the weak connection edge is lower than the edge weight of the connection edge established based on the recommendation and transaction relationship; the edge weight participates in the calculation of the recommendation and transaction probability through node features.

5. The dispatching method according to claim 1, characterized in that, The step of updating the historical bipartite graph based on the graph network update information to obtain the current bipartite graph includes: Get bipartite graph update information; Based on the node update information and node feature update information in the bipartite graph update information, adjust the salesperson nodes and their corresponding node features, as well as the customer nodes and their corresponding node features in the historical bipartite graph, to obtain the current bipartite graph.

6. The dispatching method according to claim 1, characterized in that, The step of assigning tasks to the salespersons corresponding to the salesperson nodes based on the recommended transaction probability includes: Calculate the ranking value of each salesperson based on the historical transaction information; The salespersons are sorted according to the sorting value to obtain a salesperson queue; For each salesperson in the salesperson queue, the salesperson is assigned a task based on the recommended transaction probability of the salesperson node corresponding to that salesperson.

7. A dispatching device, characterized in that, include: The information acquisition module is used to acquire historical transaction information; The bipartite graph construction module is used to construct a historical bipartite graph based on the historical transaction information; the salesperson and customer in the historical transaction information correspond to the salesperson node and customer node in the historical bipartite graph, respectively, and the salesperson node and the customer node have node characteristics; The probability prediction module is used to calculate the probability of a successful recommendation between a salesperson node and a customer node based on the node characteristics using an initial prediction model; the initial prediction model is constructed based on the GraphSAGE algorithm. The probability prediction module is also used to randomly collect the one-hop neighbor nodes and two-hop neighbor nodes of the salesperson node through the initial prediction model; Based on the node features of the one-hop neighbor node and the two-hop neighbor node, a node embedding representation of the salesperson node is generated. The node is embedded into the fully connected layer of the initial prediction model to obtain the recommendation success probability between the salesperson node and the customer node. The client node is a node among the one-hop neighbor node and the two-hop neighbor node; The model adjustment module is used to calculate the model loss based on the connection relationship between the salesperson node and the customer node in the bipartite graph and the recommendation transaction probability, and adjust the initial prediction model based on the model loss until the initial prediction model converges to obtain the prediction model. The bipartite graph update module is used to update the historical bipartite graph according to the graph network update information to obtain the current bipartite graph; The transaction prediction module is used to calculate the probability of a recommended transaction between the salesperson node to be predicted and the customer node to be predicted in the current two-part graph through the prediction model, and to assign tasks to the salesperson corresponding to the salesperson node according to the recommended transaction probability.

8. A computer device comprising a memory and a processor, the memory storing computer-readable instructions, wherein the processor, when executing the computer-readable instructions, implements the steps of the dispatching method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the dispatching method as described in any one of claims 1 to 6.

Images