Method, device and server for determining target object

Abstract

This specification provides a method, device, and server for determining a target object. Based on this method, when it is necessary to search for potential customer objects for pushing target business products, you can first obtain and construct a corresponding relationship knowledge graph based on the relationship data of the first object; then, based on the relationship knowledge graph, screen the first object to find the second object that meets the preset first requirement; then, you can call the preset potential object prediction model to process the individual data of the above-mentioned second object to obtain the prediction result of the second object; then, based on the prediction result of the second object, screen the second object to find the target object that meets the preset second requirement as the potential customer object to push the target business product. Therefore, by comprehensively utilizing the relationship data and individual data of the first object, it is possible to efficiently and accurately screen out potential customer objects with a higher probability of accepting the target business product from the first object.

Description

Technical Field

[0001] This specification belongs to the field of artificial intelligence technology, and in particular relates to a method, device, and server for determining a target object. Background Art

[0002] In many product promotion scenarios (e.g., banks promoting wealth management products), existing methods often prevent product promotion staff from fully and effectively utilizing the vast amount of customer data. Instead, they rely on a few pieces of data, combined with personal experience, to subjectively determine whether a customer is a potential customer (or potential lead). This results in significant errors and low accuracy in identifying potential customers, which in turn impacts the effectiveness of subsequent product promotion.

[0003] To address the above issues, no effective solutions have been proposed so far. Summary of the Invention

[0004] This specification provides a method, device, and server for determining a target object, so as to efficiently and accurately screen out potential customer objects with a high probability of accepting a target business product.

[0005] The embodiments of this specification provide a method for determining a target object, including:

[0006] Acquire individual data and relationship data of a plurality of first objects;

[0007] Constructing a corresponding relationship knowledge graph based on the relationship data of the plurality of first objects;

[0008] According to the relationship knowledge graph, screening out a plurality of second objects that meet a preset first requirement from the plurality of first objects;

[0009] Calling a preset potential object prediction model to process the individual data of the plurality of second objects to obtain prediction results of the second objects;

[0010] According to the prediction result of the second object, a target object that meets the preset second requirement is screened out from the plurality of second objects.

[0011] In some embodiments, the first object includes: a natural person object, and / or a legal person object.

[0012] In some embodiments, the relationship data includes at least one of the following: natural person relationship, equity relationship, position relationship, business relationship, and financial relationship.

[0013] In some embodiments, constructing a corresponding relationship knowledge graph based on the relationship data of the plurality of first objects includes:

[0014] Establishing a node according to the object identifier of the first object;

[0015] According to the relationship data between the first objects, corresponding nodes are connected using edges; and attribute values ​​of the edges are marked to obtain the relationship knowledge graph.

[0016] In some embodiments, based on the relationship knowledge graph, screening out a plurality of second objects that meet a preset first requirement from the plurality of first objects includes:

[0017] Searching the relational knowledge graph and finding a node whose object identifier matches the customer list as a starting node;

[0018] Starting from the starting node, searching for nodes connected to the starting node through edges as candidate nodes;

[0019] The first objects corresponding to the candidate nodes whose attribute values ​​meet a preset second requirement are obtained and selected as the second objects based on the attribute values ​​of the edges between the candidate nodes and the starting node.

[0020] In some embodiments, the individual data includes at least one of the following:

[0021] The first object's loan data, the first object's income data, the first object's asset data, the first object's payment record, the first object's insurance data, and the first object's registration information.

[0022] In some embodiments, calling a preset potential object prediction model to process the individual data of the plurality of second objects to obtain prediction results for the second objects includes:

[0023] Preprocessing the individual data of the second object to extract individual features of multiple dimensions of the second object;

[0024] A preset potential object prediction model is called to process individual features of multiple dimensions of the second object to obtain a prediction result of the second object.

[0025] In some embodiments, the individual characteristics of the multiple dimensions include: occupational characteristics, income characteristics, housing characteristics, vehicle characteristics, consumption characteristics, and natural person relationship characteristics.

[0026] In some embodiments, after selecting a target object that meets a preset second requirement from a plurality of second objects based on the prediction result of the second object, the method further includes:

[0027] Get the business tag of the target object;

[0028] Determine a target push strategy for the target object based on the service tag;

[0029] According to the target push strategy, link data about the target business product is pushed to the target object.

[0030] In some embodiments, the method further comprises:

[0031] Acquire individual features of multiple sample objects as sample data;

[0032] According to the business label of the sample object, the sample data is categorized to obtain labeled sample data;

[0033] Constructing an initial model and a preset loss function; wherein the preset loss function is a FocalLoss loss function;

[0034] The initial model is trained using the preset loss function and labeled sample data to obtain the preset potential object prediction model.

[0035] The embodiment of this specification also provides a device for determining a target object, including:

[0036] An acquisition module, configured to acquire individual data and relationship data of a plurality of first objects;

[0037] A construction module, configured to construct a corresponding relationship knowledge graph based on the relationship data of the plurality of first objects;

[0038] A first screening module is configured to screen out a plurality of second objects that meet a preset first requirement from the plurality of first objects according to the relationship knowledge graph;

[0039] a calling module, configured to call a preset potential object prediction model to process the individual data of the plurality of second objects to obtain prediction results of the second objects;

[0040] The second screening module is used to screen out a target object that meets a preset second requirement from a plurality of second objects according to the prediction result of the second object.

[0041] An embodiment of this specification also provides a server, including a processor and a memory for storing processor-executable instructions, wherein the processor implements the following when executing the instructions: obtaining individual data and relationship data of multiple first objects; constructing a corresponding relationship knowledge graph based on the relationship data of the multiple first objects; screening out multiple second objects that meet the preset first requirements from the multiple first objects based on the relationship knowledge graph; calling a preset potential object prediction model to process the individual data of the multiple second objects to obtain prediction results of the second objects; and screening out target objects that meet the preset second requirements from the multiple second objects based on the prediction results of the second objects.

[0042] An embodiment of this specification also provides a computer storage medium having computer instructions stored thereon, which, when executed, implement the following: obtaining individual data and relationship data of multiple first objects; constructing a corresponding relationship knowledge graph based on the relationship data of the multiple first objects; screening out multiple second objects that meet preset first requirements from the multiple first objects based on the relationship knowledge graph; calling a preset potential object prediction model to process the individual data of the multiple second objects to obtain prediction results for the second objects; and screening out target objects that meet the preset second requirements from the multiple second objects based on the prediction results for the second objects.

[0043] This specification provides a method, device and server for determining a target object. Based on this method, when it is necessary to search for potential customer objects to push target business products, you can first obtain and build a corresponding relationship knowledge graph based on the relationship data of the first object; then, based on the relationship knowledge graph, screen the first object to find the second object that meets the preset first requirement; then, you can call the preset potential object prediction model to process the individual data of the above-mentioned second object to obtain the prediction result of the second object; then, based on the prediction result of the second object, screen the second object to find the target object that meets the preset second requirement. Therefore, by comprehensively utilizing the relationship data and individual data of the first object, potential customer objects with a high probability of accepting the target business product can be efficiently and accurately screened from the first object, reducing the determination error when determining the target object. Subsequently, according to the matching strategy, information related to the target business product can be pushed to the above-mentioned target object to improve the promotion effect of the target business product. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of this specification, the following is a brief introduction to the drawings required for use in the embodiments. The drawings described below are only some of the embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0045] Figure 1 This is a schematic diagram of an embodiment of the structural composition of a system applying the target object determination method provided in the embodiments of this specification;

[0046] Figure 2 This is a flowchart of a method for determining a target object provided by an embodiment of this specification;

[0047] Figure 3 This is a schematic diagram of the structure of a server provided by an embodiment of this specification;

[0048] Figure 4This is a schematic diagram of the structure of a device for determining a target object provided by an embodiment of this specification;

[0049] Figure 5 This is a schematic diagram of an embodiment of a method for determining a target object provided by an embodiment of this specification, applied in a scenario example;

[0050] Figure 6 This is a schematic diagram of an embodiment of a method for determining a target object provided by an embodiment of this specification, applied in a scenario example;

[0051] Figure 7 This is a schematic diagram of an embodiment of a method for determining a target object provided by an embodiment of this specification, applied in a scenario example. DETAILED DESCRIPTION

[0052] To help those skilled in the art better understand the technical solutions in this specification, the following will provide a clear and complete description of the technical solutions in the embodiments of this specification, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. All other embodiments derived by those skilled in the art based on the embodiments in this specification without creative effort shall fall within the scope of protection of this specification.

[0053] The embodiment of this specification provides a method for determining a target object, which can be applied to a system including a server and a terminal device. Figure 1 The server and the terminal device can be connected via wired or wireless means to perform specific data interaction.

[0054] In this embodiment, the server may specifically include a backend server implemented on a bank's data center network platform, capable of performing functions such as data transmission and data processing. Specifically, the server may be, for example, an electronic device with data computing, storage, and network interaction capabilities. Alternatively, the server may be a software program running on the electronic device that provides support for data processing, storage, and network interaction. In this embodiment, the number of servers included in the server is not specifically limited. The server may be a single server, several servers, or a server cluster formed by several servers.

[0055] In this embodiment, the terminal device may specifically include a front-end electronic device applied to the bank's data center network platform, capable of performing functions such as data collection, data monitoring, and data transmission for the massive amount of business data accessed by the network platform. Alternatively, the terminal device may be a software application capable of running on the aforementioned electronic device. For example, it may be an app running on a server.

[0056] In this example, a bank's data center network platform receives massive amounts of business data every day. For example, data on the mortgage loan processed by user A at the bank, transfer transaction data sent by user B to user C via the bank's electronic banking service, and registration information filled out by user D when opening a linked account at the bank.

[0057] Currently, the bank plans to promote a newly launched wealth management product (which can be recorded as the target business product). To improve the promotion effect, the server can apply the target object determination method provided in the embodiments of this specification to effectively and fully utilize the massive business data accessed by the network platform to screen out customer objects with a high probability of accepting the wealth management product and a high willingness to purchase from multiple candidate customer objects (which can be recorded as first objects, including individual objects and legal person objects) as potential customer objects (which can be recorded as target objects); then, the wealth management product can be promoted to the aforementioned target objects in a targeted manner.

[0058] In specific implementation, first, the server can collect and obtain the massive business data connected to the bank's data center network platform through terminal equipment.

[0059] Then, the server can perform semantic recognition and relationship reasoning on the above business data to extract the object identifier of the first object involved in the above business data, the individual data of the first object (for example, the first object's mortgage data, the first object's car loan data, the first object's cash flow record, the first object's insurance data, etc.), and the relationship data between the first objects (for example, the natural person relationship, equity relationship, position relationship, business dealing relationship, financial relationship, etc. between the first objects).

[0060] Furthermore, the server can use the relationship data of multiple first objects to construct a corresponding relationship knowledge graph. Specifically, the server can first determine the object identifier of the first object involved in the relationship data (for example, the name of the first object, the registered mobile phone number of the first object, the account name of the first object, etc.). Then, based on the object identifier of the first object, use a circle to establish a node corresponding to the first object. The server can then use the edges used to characterize the relationship to connect the nodes corresponding to different first objects involved in the same relationship data based on the relationship data between the first objects. At the same time, according to the preset relationship annotation rules, based on the specific relationship data, the attribute values ​​of the edges (for example, the relationship type of the relationship represented by the edge, the closeness of the relationship, the duration of the relationship, and other parameters) can be annotated. In this way, the server can establish a relationship knowledge graph that meets the requirements.

[0061] The server can perform a first screening based on the relationship knowledge graph and the relationship data between the first objects, so as to screen out first objects that are relatively likely to become potential customer objects as second objects that meet the preset first requirements based on the relationships between different first objects.

[0062] Specifically, the server can first use the existing customer list with records of potential customer objects (or high-quality customer objects) to search the object identifier of the relationship knowledge image to find a node whose object identifier matches the customer list as the starting node. Then, the server can start from the starting node and search for other nodes that are directly or indirectly connected to the node through edges as candidate nodes. Then, the server can obtain the edge attribute value connecting the candidate node and the starting node. Based on the attribute value, the first object corresponding to the candidate node that has a relatively close association relationship with the starting node (that is, the candidate node that meets the preset first requirement) is screened out as the second object that meets the preset first requirement.

[0063] The server can call a pre-trained preset potential object prediction model and use the individual data of the second object to perform a second screening, so as to further screen out the first object with a relatively high probability of accepting and purchasing the target business product to be promoted from multiple first objects based on the individual characteristics of the second object, as the target object (i.e., potential customer object) that meets the preset second requirement.

[0064] Specifically, the server can first pre-process the individual data of the second object to extract individual characteristics of multiple dimensions for the second object (for example, the second object's occupational characteristics, income characteristics, housing characteristics, vehicle characteristics, consumption characteristics, natural person relationship characteristics, etc.). The server then uses the combination of the individual characteristics of the second object in multiple dimensions as model input and inputs it into a pre-trained preset potential object prediction model; and runs the model to output the prediction result of the second object. Based on the prediction result of the second object, the server can then filter out target objects from the second object that meet the preset second requirement and have a high probability of accepting and purchasing the target business product.

[0065] Finally, the server can promote target business products to target objects in a targeted manner.

[0066] Specifically, the server can first search the customer database based on the target object's object identifier to obtain the bank's business tags for the target object (for example, tags indicating the customer's investment and financial management preferences, behavioral habits, etc.), as well as other relevant customer data for the target object. Based on the target object's business tags and other relevant customer data, a matching targeted push policy can be generated for the target object. Based on this targeted push policy, link data related to the target business product can be pushed to the target object.

[0067] For example, based on the target subject's business tag, the server may discover that the target subject is relatively accustomed to using a mobile phone and that they prioritize stability and value preservation when investing and managing their finances. Based on this information, the server can generate a targeted push strategy that matches the target subject. Furthermore, based on the targeted push strategy, the server can search a pre-set promotional text library for promotional text about the target product that highlights its stability and value preservation characteristics, and use this as the targeted promotional text. The server then combines this targeted promotional text with the download link data for the target product to generate targeted promotional data about the target product for the target subject. This targeted promotional data is then sent to the target subject's mobile phone via text message. This allows the target subject to receive the targeted promotional data promptly via their mobile phone. Based on the targeted promotional text in the targeted promotional data, the target subject is more likely to develop an interest in the target product and, consequently, be more likely to download and purchase the target product by triggering the download link data in the targeted promotional data.

[0068] Through the above embodiment, potential customer objects with a high probability of accepting the target business product can be screened out from a large number of first objects in a relatively efficient and accurate manner, reducing errors in determining the target objects. Furthermore, for the above target objects, targeted promotion data related to the target business product can be generated and pushed to the target objects based on the matching targeted promotion strategy, thereby improving the promotion effect of the target business product.

[0069] See Figure 2 As shown, the embodiment of this specification provides a method for determining a target object. The method is specifically applied to the server side. When implemented, the method may include the following:

[0070] S201: Acquire individual data and relationship data of a plurality of first objects;

[0071] S202: Constructing a corresponding relationship knowledge graph based on the relationship data of the plurality of first objects;

[0072] S203: Filtering, based on the relationship knowledge graph, a plurality of second objects that meet a preset first requirement from the plurality of first objects;

[0073] S204: calling a preset potential object prediction model to process the individual data of the plurality of second objects to obtain prediction results of the second objects;

[0074] S205: Filtering a target object that meets a preset second requirement from a plurality of second objects according to the prediction result of the second object.

[0075] Through the above embodiment, two different types of data, namely the relational data and individual data of the first object, can be effectively and fully utilized to efficiently and accurately screen out potential customer objects with a higher probability of accepting the target business product from the first object as target objects, thereby reducing processing errors in the process of determining the target objects.

[0076] In some embodiments, the first object may be understood as a customer object whose potential customer is to be determined for a target business product. The target business product may be a physical product (e.g., a mobile phone, a computer, a book, etc.) or a virtual service (e.g., a financial service, an insurance service, a cleaning service, etc.).

[0077] In some embodiments, for a financial service promotion scenario (for example, a bank promoting a newly launched financial product), the first object may specifically include: a natural person object, and / or a legal person object, etc.

[0078] Of course, the first objects listed above are only for illustrative purposes. For different application scenarios, the first objects may also include other types of objects. For example, for a mobile phone promotion scenario, the first objects may specifically include: student objects, adult objects, etc.

[0079] Through the above embodiments, for the financial service promotion scenario, the target object determination method provided in the embodiments of this specification can be applied to effectively cover and perform specific judgment and determination on multiple first objects in the application scenario.

[0080] In some embodiments, the individual data may specifically include data used to reflect individual characteristics of the first object. The relationship data may specifically include data used to reflect association relationships between different first objects.

[0081] In some embodiments, for the financial service promotion scenario, the relationship data may specifically include at least one of the following: natural person relationship, equity relationship, position relationship, business relationship, financial relationship, etc.

[0082] Through the above embodiments, various types of relationship data can be comprehensively distinguished and utilized to better cover most of the association relationships in the financial service promotion scenario, thereby effectively utilizing the relationship data to determine potential customer objects.

[0083] In some embodiments, for the financial service promotion scenario, the individual data may specifically include at least one of the following: the first object's loan data, the first object's income data, the first object's asset data, the first object's payment record, the first object's insurance data, the first object's registration information, etc.

[0084] Through the above embodiments, various types of individual data can be obtained and utilized more comprehensively to better cover most of the relevant individual characteristics in the financial service promotion scenario, thereby effectively utilizing individual data to identify potential customer objects.

[0085] In some embodiments, the above-mentioned acquisition of individual data and relationship data of multiple first objects may include the following when implemented:

[0086] S1: Acquire business data related to a first object; wherein the business data may specifically include: registration information of the first object, user data of the first object, product holding records of the first object, transaction data between the first objects, asset registration data involving the first object, etc.;

[0087] S2: Perform semantic recognition and association reasoning on the business data to extract individual data of the first object and relationship data between the first objects.

[0088] In some embodiments, the above-mentioned construction of a corresponding relationship knowledge graph based on the relationship data of the plurality of first objects may include the following contents during specific implementation:

[0089] S1: Create a node according to the object identifier of the first object;

[0090] S2: Based on the relationship data between the first objects, connect the corresponding nodes using edges; and mark the attribute values ​​of the edges to obtain the relationship knowledge graph.

[0091] Through the above embodiments, the relationship data between different first objects can be more effectively utilized to construct a relationship knowledge graph that contains more comprehensive association relationships and has higher accuracy.

[0092] In some embodiments, the above-mentioned filtering out a plurality of second objects that meet a preset first requirement from the plurality of first objects according to the relationship knowledge graph may include the following steps during implementation:

[0093] S1: Search the relationship knowledge graph and find a node whose object identifier matches the customer list as the starting node;

[0094] S2: Starting from the starting node, search for nodes connected to the starting node through edges as candidate nodes;

[0095] S3: Obtain and, based on the attribute values ​​of the edges between the candidate nodes and the starting node, select first objects corresponding to the candidate nodes whose attribute values ​​meet a preset second requirement as second objects.

[0096] Through the above embodiment, the relationship data between different first objects can be effectively utilized according to the relationship knowledge graph to complete the first screening, so as to screen out the second objects that may be potential customer objects from the massive first objects.

[0097] In some embodiments, the customer list may specifically be list data that records the identifications of objects that have been determined to be potential customer objects or high-potential customer objects.

[0098] In some embodiments, for the financial service promotion scenario, the above-mentioned method of filtering out multiple second objects that meet the preset first requirement from the multiple first objects based on the relational knowledge graph may also include: based on the individual data of the first object, filtering out the first object whose personal asset data is greater than the preset data preset as a candidate object; searching the relational knowledge graph to find the node with the same object identifier as the candidate object as the starting node; and then starting from the starting node, searching for the node connected to the starting node by an edge as a candidate node; and then obtaining and filtering out the first object corresponding to the candidate node whose attribute value meets the preset second requirement as the second object based on the attribute value of the edge between the candidate node and the starting node.

[0099] In some embodiments, the above-mentioned calling of a preset potential object prediction model to process the individual data of the multiple second objects to obtain the prediction results of the second objects may include the following contents during specific implementation: preprocessing the individual data of the second objects to extract the individual features of multiple dimensions of the second objects; calling the preset potential object prediction model to process the individual features of multiple dimensions of the second objects to obtain the prediction results of the second objects.

[0100] Through the above embodiment, the individual data of the second object can be preprocessed first to extract the corresponding individual features; then the pre-trained preset potential object prediction model is called to process the above individual features to accurately determine the prediction result indicating whether the second object is a potential customer object; and then, based on the prediction result of the second object, the individual data of the second object can be fully utilized to complete the second screening, so as to further determine and screen out the potential customer object from multiple second objects as the final target object.

[0101] In some embodiments, the individual characteristics of the multiple dimensions may specifically include: occupational characteristics, income characteristics, housing characteristics, vehicle characteristics, consumption characteristics, natural person relationship characteristics, etc.

[0102] Through the above embodiments, individual characteristics of multiple different dimensions can be more comprehensively distinguished and utilized to better cover most individual characteristics in the financial service promotion scenario, and then based on the individual characteristics of the second object, individual data can be effectively used to finely identify and determine potential customer objects.

[0103] In some embodiments, the above-mentioned preprocessing of the individual data of the second object may include: performing semantic recognition on the individual data of the second object to extract multiple feature data of the second object; calling a pre-trained feature classification model to process the multiple feature data to obtain individual features of multiple dimensions of the second object.

[0104] In some embodiments, the preset potential object prediction model may specifically be an XGBoost algorithm model.

[0105] In some embodiments, XGBoost can be understood as an algorithm implementation based on GBDT, another algorithm model based on the concept of boosting ensembles. When training a model based on the XGBoost algorithm model, a feed-forward distribution algorithm can be used for greedy learning. Each iteration learns a CART number to fit the residual between the prediction results of the previous t-1 trees and the true value of the training sample. XGBoost supports parallel computing.

[0106] In some embodiments, when specifically calling a preset potential object prediction model to process the individual features of multiple dimensions of the second object, the individual features of each dimension can be first converted into corresponding feature codes according to the preset feature coding rules corresponding to the feature dimensions; and then the multiple feature codes are combined as model input and input into the preset potential object prediction model for specific processing.

[0107] In some embodiments, for the promotion of financial services, the aforementioned occupational characteristics can, to a certain extent, reflect the client's asset acquisition ability. For example, occupation A generally has a higher and more stable asset acquisition ability compared to other occupations. Prior to implementation, the following preset feature encoding rules can be designed for these occupational characteristics: 1 - Occupation A, 2 - Occupation B, 3 - Occupation C, 4 - Occupation D, and so on.

[0108] For the promotion scenario of financial services, the above-mentioned income characteristics can better assist in inferring the asset situation of the customer, for example, including: salary payment information, party fee payment information, government information (social insurance, provident fund), etc.

[0109] For the promotion scenario of financial services, the above housing characteristics can also reflect the customer's class and wealth to a certain extent, for example, including: mortgage information, reliable contact address, decoration loan information, etc.

[0110] For the promotion of financial services, the above-mentioned vehicle characteristics also reflect the important assets and consumption capacity of the customer target, for example, including: ETC registration information, driving license scanning information, car consumer loan information, etc.

[0111] When promoting wealth management services, these consumption characteristics often correlate with a customer's purchasing power, which is primarily supported by their assets. Therefore, they can also reflect the customer's asset status. Examples include average monthly credit card spending, credit card spending (luxury goods, overseas spending), children's school fees (tuition), and physical precious metal investments.

[0112] In the context of financial services promotion, the aforementioned natural person relationship features, such as the natural person relationship between a client and identified potential clients (or high-potential clients, ultra-high-net-worth clients, etc.), can also indirectly reflect the client's own asset status. Prior to implementation, the following pre-defined feature coding rules can be designed for these natural person relationship features: 11 - Category 1 relationship, 12 - Category 2 relationship, 13 - Category 3 relationship, etc.

[0113] In some embodiments, after selecting a target object that meets the preset second requirement from a plurality of second objects based on the prediction result of the second object, the method may further include the following steps when implemented:

[0114] S1: Get the business tag of the target object;

[0115] S2: Determine a target push strategy for the target object based on the service tag;

[0116] S3: Push link data about the target business product to the target object according to the target push strategy.

[0117] Through the above embodiment, after determining the target object with a higher probability of accepting the target business product from a massive number of first objects, it is also possible to determine and push the link data about the target business product to the target object in a targeted manner based on the target push strategy that matches the target object, thereby achieving better promotion effects.

[0118] In some embodiments, the method may further include the following when implemented:

[0119] S1: Obtain individual characteristics of multiple sample objects as sample data;

[0120] S2: Mark the sample data by category to obtain marked sample data;

[0121] S3: Constructing an initial model and a preset loss function; wherein the preset loss function is a FocalLoss loss function;

[0122] S4: Use the preset loss function and labeled sample data to train the initial model to obtain the preset potential object prediction model.

[0123] Through the above embodiments, sample data can be used, and in combination with a preset loss function based on the FocalLoss loss function, a preset potential object prediction model with high accuracy and good effect can be efficiently trained.

[0124] In some embodiments, the aforementioned classification labeling of the sample data to obtain the labeled sample data may include, in specific implementation, setting corresponding category labels on the sample data according to the category of the sample data (e.g., whether it is a high-potential customer object or a non-high-potential customer object; or whether it is a customer object with successful promotion or a customer object with failed promotion), thereby obtaining the labeled sample data. For example, a category label representing a positive sample is set on the sample data of a high-potential customer object or a customer object with successful promotion, and a category label representing a negative sample is set on the sample data of a non-high-potential customer object or a customer object with failed promotion, thereby obtaining the labeled sample data.

[0125] After obtaining the labeled sample data, the labeled sample data can also be split into a training set and a test set according to a preset training ratio (for example, 9:1), so that the training set can be used to train the model first, and then the test set can be used to test the model to obtain a preset potential object prediction model that meets the requirements.

[0126] In some embodiments, before using the labeled sample data for model training, the individual features of the sample objects can be first tested for numerical type to find individual features that are numerically continuous (for example, payroll information, mortgage information, average monthly credit card spending, etc.); then the histogram algorithm can be used to discretize the above-mentioned numerically continuous individual features into k discrete features, and at the same time construct a histogram with a width of k (containing k bins) for information statistics. In this way, when the model is subsequently trained, the information counted by the above-mentioned histogram can be directly used for training. There is no need to traverse all individual features. It is only necessary to traverse the k bins in the histogram to quickly find the best splitting point, thereby greatly improving the computational efficiency when splitting the node, reducing the variance of the model to a certain extent, enhancing the robustness of the model, and improving the efficiency of model training.

[0127] In some embodiments, the preset loss function may be a FocalLoss loss function. Based on the preset loss function, the FocalLoss loss function may be effectively used to set different weight coefficients for different categories.

[0128] In some embodiments, when training a preset potential object prediction model, the objective function of the model training (including a preset loss function and a regularization term) can be constructed in the following manner based on the XGBoost algorithm concept:

[0129]

[0130] Among them, Obj represents the function value of the objective function, l represents the preset loss function, Ω represents the regularization term, and f represents the model function. is the i-th sample object x i The predicted value, y i is the i-th sample object x i The true value of .

[0131] The above preset loss function can be specifically expressed as follows:

[0132]

[0133] Among them, α i Represents the weight coefficient set based on the category, and the value range is between 0 and 1.

[0134] By introducing and utilizing the above-mentioned preset loss function containing category-based weight coefficients to construct an objective function for model training, different weights can be assigned to different categories, thereby more effectively avoiding the problem that the loss of sample categories with a high proportion dominates the training process, resulting in the trained model paying special attention to the sample categories with a high proportion, thereby improving the accuracy of model training.

[0135] Furthermore, since XGBoost is an additive model, the prediction value (or prediction score) can be the cumulative sum of the scores of each tree, as shown below:

[0136] Then sum the complexity of all k trees and add it to the objective function as a regularization term:

[0137]

[0138] Where J represents the number of leaf nodes, and w represents the optimal solution for the j-th leaf node.

[0139] By introducing and utilizing regularization terms to construct the objective function for model training, overfitting can be effectively prevented and the training effect of the model can be further improved.

[0140] As can be seen from the above, based on the target object determination method provided in the embodiment of this specification, when it is necessary to search for potential customer objects to push target business products, you can first obtain and build a corresponding relationship knowledge graph based on the relationship data of the first object; based on the relationship knowledge graph, the first object is screened to find the second object that meets the preset first requirement; then, the preset potential object prediction model can be called to process the individual data of the above-mentioned second object to obtain the prediction result of the second object; then, based on the prediction result of the second object, the second object is screened to find the target object that meets the preset second requirement. Therefore, by comprehensively utilizing the relationship data and individual data of the first object, potential customer objects with a high probability of accepting the target business product can be efficiently and accurately screened from the first object. Then, for the above-mentioned target object, information related to the target business product can be pushed according to the matching strategy to improve the promotion effect of the target business product.

[0141] An embodiment of this specification also provides a server, including a processor and a memory for storing processor executable instructions. When the processor is implemented, it can perform the following steps according to the instructions: obtain individual data and relationship data of multiple first objects; construct a corresponding relationship knowledge graph based on the relationship data of the multiple first objects; based on the relationship knowledge graph, filter out multiple second objects that meet the preset first requirements from the multiple first objects; call a preset potential object prediction model to process the individual data of the multiple second objects to obtain prediction results of the second objects; based on the prediction results of the second objects, filter out target objects that meet the preset second requirements from the multiple second objects.

[0142] In order to complete the above instructions more accurately, refer to Figure 3 As shown, the embodiment of this specification also provides another specific server, wherein the server includes a network communication port 301, a processor 302 and a memory 303, and the above structures are connected through internal cables so that each structure can perform specific data interaction.

[0143] The network communication port 301 may be used to obtain individual data and relationship data of a plurality of first objects.

[0144] The processor 302 can be specifically used to construct a corresponding relational knowledge graph based on the relational data of the multiple first objects; based on the relational knowledge graph, screen out multiple second objects that meet the preset first requirements from the multiple first objects; call a preset potential object prediction model to process the individual data of the multiple second objects to obtain prediction results of the second objects; and based on the prediction results of the second objects, screen out a target object that meets the preset second requirements from the multiple second objects.

[0145] The memory 303 may be specifically used to store corresponding instruction programs.

[0146] In this embodiment, the network communication port 301 can be a virtual port that is bound to different communication protocols, thereby being capable of sending or receiving different data. For example, the network communication port can be a port responsible for web data communication, a port responsible for FTP data communication, or a port responsible for email data communication. Furthermore, the network communication port can also be a physical communication interface or communication chip. For example, it can be a wireless mobile network communication chip, such as GSM or CDMA; it can also be a Wi-Fi chip; or it can be a Bluetooth chip.

[0147] In this embodiment, the processor 302 can be implemented in any suitable manner. For example, the processor can take the form of a microprocessor or a processor and a computer-readable medium storing computer-readable program code (such as software or firmware) executable by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, an embedded microcontroller, etc. This specification is not limited to this.

[0148] In this embodiment, the memory 303 may include multiple levels. In a digital system, anything that can store binary data can be a memory. In an integrated circuit, a circuit with a storage function that does not have a physical form is also called a memory, such as RAM, FIFO, etc. In a system, a storage device with a physical form is also called a memory, such as a memory stick, TF card, etc.

[0149] An embodiment of this specification also provides a computer storage medium based on the above-mentioned target object determination method, wherein the computer storage medium stores computer program instructions, which, when executed, implement the following: obtaining individual data and relationship data of multiple first objects; constructing a corresponding relationship knowledge graph based on the relationship data of the multiple first objects; screening out multiple second objects that meet the preset first requirements from the multiple first objects based on the relationship knowledge graph; calling a preset potential object prediction model to process the individual data of the multiple second objects to obtain prediction results for the second objects; and screening out target objects that meet the preset second requirements from the multiple second objects based on the prediction results of the second objects.

[0150] In this embodiment, the storage medium includes, but is not limited to, random access memory (RAM), read-only memory (ROM), cache, hard disk drive (HDD), or memory card. The memory can be used to store computer program instructions. The network communication unit can be an interface configured in accordance with the standards specified by the communication protocol for network connection communication.

[0151] In this embodiment, the functions and effects specifically implemented by the program instructions stored in the computer storage medium can be explained in comparison with other implementations and will not be repeated here.

[0152] See Figure 4 As shown, at the software level, the embodiment of this specification further provides a device for determining a target object, which may specifically include the following structural modules:

[0153] An acquisition module 401 may be specifically configured to acquire individual data and relationship data of a plurality of first objects;

[0154] A construction module 402 may be specifically configured to construct a corresponding relationship knowledge graph based on the relationship data of the plurality of first objects;

[0155] A first screening module 403 may be specifically configured to screen out a plurality of second objects that meet a preset first requirement from the plurality of first objects according to the relationship knowledge graph;

[0156] The calling module 404 may be specifically configured to call a preset potential object prediction model to process the individual data of the plurality of second objects to obtain prediction results of the second objects;

[0157] The second screening module 405 may be specifically configured to screen out a target object that meets a preset second requirement from a plurality of second objects according to the prediction result of the second object.

[0158] It should be noted that the units, devices or modules described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. For the convenience of description, the above devices are described in terms of functions and are divided into various modules and described separately. Of course, when implementing this specification, the functions of each module can be implemented in the same or multiple software and / or hardware, or the module that implements the same function can be implemented by a combination of multiple sub-modules or sub-units. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0159] As can be seen from the above, the target object determination device provided in the embodiments of this specification, by comprehensively utilizing the relationship data and individual data of the first object, efficiently and accurately screens out potential customer objects with a high probability of accepting the target business product from the first object.

[0160] In a specific scenario example, the target object determination method provided in the embodiment of this specification can be applied to explore target customers based on the association relationship and conduct corresponding marketing. Figure 5 As shown, the following steps are included.

[0161] Step S1: Acquire and determine multiple implicit relationships (e.g., first object relationship data) based on initial business data. A knowledge graph (e.g., a relationship knowledge graph) is constructed using relationships such as natural person relationships, equity relationships, position relationships, business dealings relationships, and funding relationships. This knowledge graph can be used to explore various implicit relationships and identify customers who are temporarily characterized as having low value and low assets (e.g., total personal assets of less than 200,000 yuan), but whose actual assets are estimated to be high based on the associated data (e.g., selecting second objects that meet a preset first requirement from a large number of first objects).

[0162] Step S2: Based on the list of low-asset potential customers, potential target customers are predicted based on the six dimensions that are most closely related to funds (including: occupation, income, housing, car, consumption, and natural person relationship) using the XGBoost algorithm model (for example, a preset potential object prediction model) according to the characteristics of the above six dimensions (for example, individual characteristics).

[0163] In step S3, based on the model prediction results and combined with traditional customer star rating labels, the comprehensive situation of the customer can be judged to provide strong decision support for the customer manager's precision marketing (for example, determining the target object and promoting the target business product to the target object).

[0164] In this scenario example, during specific implementation, legal person-to-individual and person-to-individual relationship data can be used, with customer entities as "points" and the "relationships" between entities as "edges" to construct a knowledge graph to mine low-asset potential customers. For example, by combining company information (company settlement accounts, listed company information) with industrial and commercial registration information (equity relationships, executive relationships) or capital flow relationships, potential customers can be accurately identified; or starting with private bank ultra-high net worth customers, target customers can be mined based on graph relationships such as kinship, credit card primary and secondary card relationships, and corporate relationships. Kinship relationships can be determined based on other information such as personal registration information, co-payers of mortgages, loan guarantee information, credit card contacts, same credit card addresses, same contact addresses, insurance beneficiaries, fixed transfer accounts, close financial contacts, and government information (marriage).

[0165] In this scenario example, the XGBoost algorithm model can be used to predict potential target customers for these target customer groups based on basic customer information such as occupation, income, housing, cars, consumption, and natural person relationships.

[0166] Among them, Occupation: A customer's occupation determines their ability to acquire assets, especially for key corporate managers. We encode common occupations, such as 1-A Occupation, 2-B Occupation, 3-C Occupation, 4-D Occupation, etc., and use them as features for XGBoost model learning. Income: A customer's income can directly infer their assets, including payroll information, party dues payment information, and government information (social security, provident fund). Housing: The different housing conditions of domestic customers reflect their class and wealth, including mortgage information, reliable contact addresses, and renovation loan information. Cars: Cars are important assets and a reflection of a customer's spending power. This includes ETC registration information, driving license scans, and car loan information. Consumption: Consumption reflects a customer's purchasing power, which is supported by their financial resources. This includes average monthly credit card spending, credit card spending (luxury goods, overseas spending), children's school fees (tuition), and physical precious metal investment. Natural person relationships: Based on the associations obtained in step S1, the relationship between the predicted client and the private bank's ultra-high net worth clients is coded, such as 11 - Category 1 relationship, 12 - Category 2 relationship, 13 - Category 3 relationship, etc. Other content: Former high-net-worth clients who have lost due to various reasons can be used as tracking basis.

[0167] In this example scenario, see Figure 6 As shown, the corresponding system can be used to train a potential target customer prediction model based on the XGBoost algorithm. The system can specifically include the following structure:

[0168] Training sample unit 601 and test sample unit 602: In this scenario, known private banking clients and high-net-worth clients can be used as positive samples, while cases of clients whose early marketing efforts by grassroots account managers failed can be used as negative samples. Basic information such as the client's occupation, income, housing, vehicle, consumption, and natural person relationships can be used as sample features. The sample data is divided into training and test samples in a ratio of 9:1.

[0169] Data Preprocessing Unit 603: For continuous feature values ​​(such as payroll information, mortgage information, average monthly credit card spending, etc.), we use a histogram algorithm to discretize them into k discrete features and construct a k-width histogram for statistical information (containing k bins). This preprocessing method eliminates the need to traverse the data during model training; instead, the optimal split point can be found by traversing k bins. This greatly improves the computational efficiency of node splitting, reduces the model variance to a certain extent, and enhances the model's robustness.

[0170] Model training unit 604: In this example scenario, the XGBoost algorithm can be used to train the potential target customer prediction model. XGBoost is an engineering implementation of GBDT, an algorithm model based on the boosting ensemble concept. During training, the feedforward distribution algorithm performs greedy learning. Each iteration learns a CART number to fit the residual between the prediction results of the previous t-1 trees and the true value of the training sample. XGBoost shares the basic concept of GBDT with some optimizations, such as default missing value handling, the addition of second-order derivative information, regularization terms, column sampling, and parallel computing.

[0171] Among them, the objective function of XGBoost consists of two parts: loss function and regularization term, which are defined as follows:

[0172]

[0173] Where l represents the loss function. The above loss function based on logistic regression can be expressed as:

[0174]

[0175] in, is the i-th sample x iThe reason is that when preparing the sample data set, known private banking customers and high-net-worth customers are used as positive samples, and customer cases where grassroots account managers failed in marketing in the early stage are used as negative samples. There is a large gap between the proportions of positive and negative samples, which leads to the loss of sample categories with a high proportion dominating the training process. The trained model will pay special attention to the sample categories with a high proportion, resulting in a decrease in the accuracy of the model.

[0176] Therefore, in this scenario example, FocalLoss can be used as the loss function to give different weights to each category:

[0177]

[0178] Among them, α i is the category weight, and its value range is between 0 and 1.

[0179] Since XGBoost is an additive model, the prediction score is the cumulative sum of the scores of each tree:

[0180]

[0181] The complexity of all k trees is then summed and added to the objective function as a regularization term to prevent overfitting:

[0182]

[0183] Where J is the number of leaf nodes and w is the optimal solution for the j-th leaf node.

[0184] Model evaluation unit 605: Use the test sample to evaluate the trained model. When the model iteration outputs the model evaluation result that meets the prediction requirements, the training ends.

[0185] In this scenario example, as the business is used, the accumulated marketing results data and the original information data in the database can be used to retrain the potential target customer prediction model to improve the model accuracy.

[0186] For further information, see Figure 7 As shown, a marketing system based on relationship-based target customer exploration can be built. Specifically, the system may include the following structure:

[0187] 701 Legal Entity Client List Unit: Obtain the list of company settlement accounts and listed companies;

[0188] 702 Legal Person Associated Individual Exploration Unit (which may include equity relationship units, position relationship units, and capital relationship units, etc.): This unit abstracts entity objects such as corporate customers and individual customers into "points" and abstracts the relationships between entities such as business registration information (equity relationships, executive relationships) or capital flows into "edges." This generates a relationship network, which is used to explore potential individual customers based on the "relationship" dimension.

[0189] 703 High-value individual customer list unit: can obtain the list of private banking customers and high-net-worth customers;

[0190] 704 Individual-Related Individual Exploration Unit (including natural person relationship unit, capital relationship unit, policyholder relationship unit, etc.): Individual customers are simply abstracted as "points" and the relationships between entities such as kinship (including personal registration information, mortgage co-payers, loan guarantee information, credit card contacts, same credit card address, same contact address, insurance beneficiary, fixed transfer account, close financial contacts, government information (marriage)), credit card primary and secondary card relationships, and corporate relationship relationships are abstracted as "edges". This generates a relationship network and explores individual potential customers based on the "relationship" dimension.

[0191] 705 Potential Customer List Unit: Performs a personal asset query on the list of potential customers obtained by the Legal Person Related Individual Exploration Unit and the Individual Related Individual Exploration Unit. If the customer's net assets are less than 200,000 yuan, an implicit star rating assessment is performed.

[0192] 706 Individual Customer Basic Information Unit: Contains individual customer occupation unit, income unit, housing unit, car unit, consumption unit, natural person relationship unit, etc.

[0193] 707 Target Potential Customer Prediction Model: Used to predict potential target customers for target customer groups based on individual customer basic information features and using the XGBoost algorithm model;

[0194] 708 Marketing Personal Customer List Unit: Determine the comprehensive situation of customers based on the model prediction results and the traditional customer star rating, and obtain a list of customers to be marketed.

[0195] Through the above scenario examples, it can be verified that the target object determination method provided by the embodiments of this specification can break through the limitations of existing relational databases and more efficiently, accurately, and quickly mine low-asset potential customers from massive amounts of data. At the same time, it effectively utilizes relationships such as natural person relationships, equity relationships, job relationships, business relationships, and financial relationships to construct a knowledge graph to explore various implicit relationships, and mine customers who temporarily appear to be low-value and low-asset, but are predicted to be valuable customers based on the associated data, thus narrowing the scope of customers to be analyzed. Furthermore, based on six potential analysis dimensions: occupation, income, housing, automobiles, consumption, natural person relationships, and other basic customer information, an improved XGBoost algorithm can be used to construct a potential target customer prediction model, predict these target customer groups, and use the prediction results as the main marketing entry point. Based on the above method and system, it can eliminate the traditional clutter caused by piling up information from various dimensions for review, and also strengthen the summary and induction of the deeper underlying situation behind the customer. In addition, by combining customer assets with model prediction results to judge the customer's comprehensive situation, it facilitates accurate identification and marketing services for account managers.

[0196] Although this specification provides the method operation steps as described in the embodiments or flow charts, more or fewer operation steps may be included based on conventional or non-creative means. The order of steps listed in the embodiments is only one way of executing the order of many steps and does not represent the only execution order. When the device or client product in practice is executed, it can be executed in sequence or in parallel according to the method shown in the embodiments or the drawings (for example, a parallel processor or a multi-threaded processing environment, or even a distributed data processing environment). The term "comprise", "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, product or device including a series of elements includes not only those elements, but also includes other elements that are not explicitly listed, or also includes elements inherent to such process, method, product or device. In the absence of more restrictions, it is not excluded that there are other identical or equivalent elements in the process, method, product or device including the elements. Words such as first and second are used to represent names and do not represent any particular order.

[0197] Those skilled in the art will also appreciate that, in addition to implementing the controller in pure computer-readable program code, it is entirely possible to implement the same functionality by logically programming the method steps in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, embedded microcontrollers, and the like. Therefore, such a controller can be considered a hardware component, and the devices included therein for implementing various functions can also be considered structures within the hardware component. Alternatively, the devices for implementing various functions can be considered both software modules implementing the method and structures within the hardware component.

[0198] This specification may be described in the general context of computer-executable instructions, such as program modules, executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, classes, and the like that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media, including storage devices.

[0199] Through the description of the above embodiments, it can be seen that those skilled in the art can clearly understand that this specification can be implemented by means of software plus the necessary general hardware platform. Based on this understanding, the technical solution of this specification can essentially be embodied in the form of a software product. This computer software product can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a mobile terminal, a server, or a network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of this specification.

[0200] The various embodiments in this specification are described in a progressive manner. References to the common or similar parts of the various embodiments are sufficient. Each embodiment focuses on the differences from the other embodiments. This specification can be used in a variety of general-purpose or specialized computer system environments or configurations. For example, personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable electronic devices, network PCs, minicomputers, mainframe computers, and distributed computing environments that include any of the above systems or devices.

[0201] Although the present specification has been described through embodiments, those skilled in the art will appreciate that there are many modifications and variations to the present specification without departing from the spirit of the present specification. It is intended that the appended claims include these modifications and variations without departing from the spirit of the present specification.

Claims

1. A method for determining a target object, characterized in that: include: Obtaining individual data and relationship data of a plurality of first subjects; wherein the individual data includes at least one of the following: loan data of the first subjects, income data of the first subjects, asset data of the first subjects, payment records of the first subjects, insurance data of the first subjects, and registration information of the first subjects; and the relationship data includes at least one of the following: natural person relationships, equity relationships, employment relationships, business relationships, and financial relationships; Constructing a corresponding relationship knowledge graph based on the relationship data of the plurality of first objects; According to the relational knowledge graph, a plurality of second objects meeting a preset first requirement are screened out from the plurality of first objects; the method comprising: based on the individual data of the first objects, screening out first objects whose personal asset data is greater than a preset data threshold as candidate objects; searching the relational knowledge graph to find a node with the same object identifier as the candidate object as a starting node; starting from the starting node, searching for nodes connected to the starting node via edges as candidate nodes; obtaining and screening out the second objects based on attribute values ​​of the edges between the candidate nodes and the starting node; Calling a preset potential object prediction model to process the individual data of the plurality of second objects to obtain prediction results for the second objects; the preset potential object prediction model is obtained in the following manner: performing numerical type detection on individual features of the sample objects obtained based on the sample data to determine individual features of a continuous numerical value type; discretizing the individual features of the continuous numerical value type into k discrete features, constructing a histogram with a width of k, and performing information statistics; and training the preset potential object prediction model based on the statistical information of the histogram; According to the prediction result of the second object, a target object that meets the preset second requirement is screened out from the plurality of second objects.

2. The method according to claim 1, characterized in that The first object includes: a natural person object and / or a legal person object.

3. The method according to claim 1, characterized in that Constructing a corresponding relationship knowledge graph based on the relationship data of the plurality of first objects includes: Establishing a node according to the object identifier of the first object; According to the relationship data between the first objects, corresponding nodes are connected using edges; and attribute values ​​of the edges are marked to obtain the relationship knowledge graph.

4. The method according to claim 3, characterized in that According to the relationship knowledge graph, a plurality of second objects meeting a preset first requirement are screened out from the plurality of first objects, including: Searching the relational knowledge graph and finding a node whose object identifier matches the customer list as a starting node; Starting from the starting node, searching for nodes connected to the starting node through edges as candidate nodes; The first objects corresponding to the candidate nodes whose attribute values ​​meet a preset second requirement are obtained and selected as the second objects based on the attribute values ​​of the edges between the candidate nodes and the starting node.

5. The method according to claim 1, wherein Calling a preset potential object prediction model to process the individual data of the plurality of second objects to obtain prediction results of the second objects includes: Preprocessing the individual data of the second object to extract individual features of multiple dimensions of the second object; A preset potential object prediction model is called to process individual features of multiple dimensions of the second object to obtain a prediction result of the second object.

6. The method according to claim 5, characterized in that The individual characteristics of the multiple dimensions include: occupational characteristics, income characteristics, housing characteristics, vehicle characteristics, consumption characteristics, and natural person relationship characteristics.

7. The method according to claim 1, characterized in that After selecting a target object that meets a preset second requirement from a plurality of second objects according to the prediction result of the second object, the method further includes: Get the business tag of the target object; Determine a target push strategy for the target object based on the service tag; According to the target push strategy, link data about the target business product is pushed to the target object.

8. The method according to claim 1, characterized in that The method further comprises: Acquire individual features of multiple sample objects as sample data; Marking the sample data by category to obtain labeled sample data; Constructing an initial model and a preset loss function; wherein the preset loss function is a FocalLoss loss function; The initial model is trained using the preset loss function and labeled sample data to obtain the preset potential object prediction model.

9. A device for determining a target object, characterized in that: include: an acquisition module, configured to acquire individual data and relationship data of a plurality of first subjects; wherein the individual data includes at least one of the following: loan data of the first subjects, income data of the first subjects, asset data of the first subjects, payment records of the first subjects, insurance data of the first subjects, and registration information of the first subjects; and the relationship data includes at least one of the following: natural person relationships, equity relationships, employment relationships, business relationships, and financial relationships; A construction module, configured to construct a corresponding relationship knowledge graph based on the relationship data of the plurality of first objects; A first screening module is configured to screen out a plurality of second objects that meet a preset first requirement from the plurality of first objects according to the relationship knowledge graph; a calling module, configured to call a preset potential object prediction model to process the individual data of the plurality of second objects to obtain prediction results for the second objects; the preset potential object prediction model is obtained in the following manner: performing numerical type detection on individual features of the sample objects obtained based on the sample data to determine individual features of a continuous numerical type; discretizing the individual features of the continuous numerical type into k discrete features, constructing a histogram with a width of k, and performing information statistics; and training the preset potential object prediction model based on the statistical information of the histogram; a second screening module, configured to screen out a target object that meets a preset second requirement from a plurality of second objects according to the prediction result of the second object; Among them, the first screening module is specifically used to: based on the individual data of the first object, screen out the first object whose personal asset data is greater than a preset data threshold as a candidate object; search the relationship knowledge graph and find the node with the same object identifier as the candidate object as the starting node; starting from the starting node, search for the node connected to the starting node through an edge as the candidate node; obtain and screen out the second object based on the attribute value of the edge between the candidate node and the starting node.

10. A server, characterized in that: The method comprises a processor and a memory for storing processor-executable instructions, wherein the processor implements the steps of the method according to any one of claims 1 to 8 when executing the instructions.

11. A computer storage medium, characterized in that Computer instructions are stored thereon, and when the instructions are executed, the steps of the method according to any one of claims 1 to 8 are implemented.

Images