Object value prediction method and device, computer device, and storage medium

Abstract

The application relates to an object value prediction method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: clustering object features of each to-be-predicted object in a to-be-predicted object set, determining a first prediction value based on resource residual peak values of each to-be-predicted object in each object group; inputting the object features of each to-be-predicted object in the to-be-predicted object set into a regression model respectively to obtain a second prediction value; for each to-be-predicted object in the to-be-predicted object set, determining a third prediction value based on multi-dimensional relationship data between the current to-be-predicted object and each first object; determining a fourth prediction value based on a resource residual peak value of the current to-be-predicted object; and determining an object value of each to-be-predicted object based on the first prediction value, the second prediction value, the third prediction value and the fourth prediction value. The method can improve the object value prediction accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to an object value prediction method and device, computer equipment, a storage medium and a computer program product. BACKGROUND

[0002] Most modern enterprises have a large number of business objects, and hierarchical interaction of business objects with different values is important for improving the overall business level of the enterprise. In actual business, a large part of the business objects have low activity, and the value of the business objects cannot be directly obtained, so value prediction is needed.

[0003] In the existing object value prediction method, the value indicators of the objects are usually formed by artificial business experts according to business expert rules. However, this object value prediction method can only classify according to the data already exhibited by the customer, and cannot predict the potential value of the object, so there is a problem of low object value prediction accuracy. SUMMARY

[0004] Therefore, it is necessary to provide an object value prediction method, device, computer equipment, computer readable storage medium and computer program product capable of improving the accuracy of object value prediction to solve the problem of low accuracy of the traditional object value prediction method.

[0005] In a first aspect, the present application provides an object value prediction method. The method comprises:

[0006] Obtaining object features of each to-be-predicted object in a to-be-predicted object set;

[0007] Clustering the object features of each to-be-predicted object to divide each to-be-predicted object into a plurality of object groups, and determining a first predicted value of each to-be-predicted object in each object group based on a resource remaining peak value of each to-be-predicted object in the first object group within a first preset time period;

[0008] Inputting the object features of each to-be-predicted object in the to-be-predicted object set into a regression model respectively to obtain a second predicted value of each to-be-predicted object in the to-be-predicted object set;

[0009] For each to-be-predicted object in the to-be-predicted object set, determining multi-dimensional relationship data between the current to-be-predicted object and each first object in the to-be-predicted object set except the current to-be-predicted object, and determining a third predicted value of the current to-be-predicted object based on the multi-dimensional relationship data;

[0010] For each to-be-predicted object in the to-be-predicted object set, determining a fourth predicted value of the current to-be-predicted object based on a resource remaining peak value of the current to-be-predicted object within a second preset time period;

[0011] Determine an object value of each object to be predicted based on the first prediction value, the second prediction value, the third prediction value and the fourth prediction value of each object to be predicted in the set of objects to be predicted.

[0012] In one of the embodiments, the training step of the regression model comprises:

[0013] Obtain the object features of the training objects and the value labels of the training objects;

[0014] Input the object features of the training objects into the current training model to obtain the predicted values of the training objects, judge whether the current training model meets the preset stopping condition based on the predicted values and the value labels, if yes, take the current training model as the regression model, if not, continue to train the current training model until the current training model meets the preset stopping condition.

[0015] In one of the embodiments, the third prediction value of the current object to be predicted is determined based on the multi-dimensional relationship data, comprising:

[0016] Determine a sub-graph corresponding to the current object to be predicted on the knowledge graph of the set of objects to be predicted;

[0017] For each first object in the set of objects to be predicted except the current object to be predicted, perform a weighted average on the multi-dimensional relationship data of the current object to be predicted and the current first object to obtain an intimate distance between the current object to be predicted and the current first object;

[0018] According to the intimate distance between the current object to be predicted and the current first object, obtain a propagation weight between the current object to be predicted and the current first object;

[0019] Based on the propagation weights between the current object to be predicted and each first object, determine the third prediction value of the current object to be predicted in the sub-graph corresponding to the current object to be predicted.

[0020] In one of the embodiments, the sub-graph corresponding to the current object to be predicted is determined on the knowledge graph of the set of objects to be predicted, comprising:

[0021] Construct the knowledge graph of the set of objects to be predicted, wherein each object to be predicted in the set of objects to be predicted corresponds to a node in the knowledge graph;

[0022] Select relationship nodes within a preset level relationship range of the current object to be predicted in the knowledge graph with the node corresponding to the current object to be predicted as the center;

[0023] Determine the sub-graph corresponding to the current object to be predicted based on the node corresponding to the current object to be predicted and the relationship nodes.

[0024] In one of the embodiments, based on the propagation weight between the current to-be-predicted object and each first object, the third prediction value of the current to-be-predicted object is determined in the sub-graph corresponding to the current to-be-predicted object, including:

[0025] In the sub-graph corresponding to the current to-be-predicted object, the shortest path from each relationship node to the node corresponding to the current to-be-predicted object is determined, and the target node passed through by the shortest path;

[0026] The propagation weight between adjacent two target nodes in the target nodes is obtained;

[0027] The resource residual peak value of each relationship node corresponding to-be-predicted object in the third preset time period is obtained;

[0028] Based on the propagation weight between adjacent two target nodes in the target nodes and the resource residual peak value of each relationship node corresponding to-be-predicted object in the third preset time period, the value influence of each relationship node corresponding to-be-predicted object on the current to-be-predicted object is determined;

[0029] The value influence of each relationship node corresponding to-be-predicted object on the current to-be-predicted object is summed up to determine the third prediction value of the current to-be-predicted object.

[0030] In one of the embodiments, based on the first prediction value, the second prediction value, the third prediction value and the fourth prediction value of each to-be-predicted object in the to-be-predicted object set, the object value of each to-be-predicted object is determined, including:

[0031] For each to-be-predicted object in the to-be-predicted object set, the maximum value is selected from the first prediction value, the second prediction value, the third prediction value and the fourth prediction value of the current to-be-predicted object, and the maximum value is taken as the object value of the current to-be-predicted object.

[0032] In a second aspect, the application further provides an object value prediction device. The device comprises:

[0033] A feature acquisition module is configured to acquire the object features of each to-be-predicted object in the to-be-predicted object set;

[0034] A first value acquisition module is configured to cluster the object features of each to-be-predicted object to divide the to-be-predicted objects into a plurality of object groups, and determine the first prediction value of each to-be-predicted object in each object group based on the resource residual peak value of each to-be-predicted object in the first preset time period;

[0035] A second value acquisition module is configured to input the object features of each to-be-predicted object in the to-be-predicted object set into a regression model to obtain the second prediction value of each to-be-predicted object in the to-be-predicted object set;

[0036] The third value obtaining module is configured to determine, for each to-be-predicted object in the set of to-be-predicted objects, multi-dimensional relationship data between the current to-be-predicted object and each first object in the set of to-be-predicted objects except the current to-be-predicted object, and determine a third prediction value of the current to-be-predicted object based on the multi-dimensional relationship data.

[0037] The fourth value obtaining module is configured to determine, for each to-be-predicted object in the set of to-be-predicted objects, a fourth prediction value of the current to-be-predicted object based on a resource remaining peak value of the current to-be-predicted object in the second preset time period.

[0038] The value determining module is configured to determine an object value of each to-be-predicted object based on the first prediction value, the second prediction value, the third prediction value, and the fourth prediction value of each to-be-predicted object in the set of to-be-predicted objects.

[0039] In a third aspect, the present application further provides a computer device. The computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0040] Obtaining object features of each to-be-predicted object in the set of to-be-predicted objects;

[0041] Clustering the object features of each to-be-predicted object to divide the to-be-predicted objects into a plurality of object groups, and determining a first prediction value of each to-be-predicted object in each object group based on a resource remaining peak value of each to-be-predicted object in the first preset time period;

[0042] Inputting the object features of each to-be-predicted object in the set of to-be-predicted objects into a regression model respectively to obtain a second prediction value of each to-be-predicted object in the set of to-be-predicted objects;

[0043] Determining, for each to-be-predicted object in the set of to-be-predicted objects, multi-dimensional relationship data between the current to-be-predicted object and each first object in the set of to-be-predicted objects except the current to-be-predicted object, and determining a third prediction value of the current to-be-predicted object based on the multi-dimensional relationship data.

[0044] Determining, for each to-be-predicted object in the set of to-be-predicted objects, a fourth prediction value of the current to-be-predicted object based on a resource remaining peak value of the current to-be-predicted object in the second preset time period.

[0045] Determining an object value of each to-be-predicted object based on the first prediction value, the second prediction value, the third prediction value, and the fourth prediction value of each to-be-predicted object in the set of to-be-predicted objects.

[0046] In a fourth aspect, the present application also provides a computer readable storage medium. The computer readable storage medium has a computer program stored thereon, and the computer program, when executed by a processor, implements the following steps:

[0047] obtaining object features of each to-be-predicted object in the set of to-be-predicted objects;

[0048] grouping the to-be-predicted objects into a plurality of object groups by clustering the object features of each to-be-predicted object, and determining a first prediction value of each to-be-predicted object in each object group based on a resource remaining peak value of each to-be-predicted object in the first preset time period;

[0049] inputting the object features of each to-be-predicted object in the set of to-be-predicted objects into a regression model respectively to obtain a second prediction value of each to-be-predicted object in the set of to-be-predicted objects;

[0050] for each to-be-predicted object in the set of to-be-predicted objects, determining multi-dimensional relationship data between the current to-be-predicted object and each first object in the set of to-be-predicted objects except the current to-be-predicted object, and determining a third prediction value of the current to-be-predicted object based on the multi-dimensional relationship data;

[0051] for each to-be-predicted object in the set of to-be-predicted objects, determining a fourth prediction value of the current to-be-predicted object based on a resource remaining peak value of the current to-be-predicted object in the second preset time period;

[0052] determining an object value of each to-be-predicted object based on the first prediction value, the second prediction value, the third prediction value and the fourth prediction value of each to-be-predicted object in the set of to-be-predicted objects.

[0053] In a fifth aspect, the present application also provides a computer program product. The computer program product comprises a computer program, and the computer program, when executed by a processor, implements the following steps:

[0054] obtaining object features of each to-be-predicted object in the set of to-be-predicted objects;

[0055] grouping the to-be-predicted objects into a plurality of object groups by clustering the object features of each to-be-predicted object, and determining a first prediction value of each to-be-predicted object in each object group based on a resource remaining peak value of each to-be-predicted object in the first preset time period;

[0056] inputting the object features of each to-be-predicted object in the set of to-be-predicted objects into a regression model respectively to obtain a second prediction value of each to-be-predicted object in the set of to-be-predicted objects;

[0057] For each to-be-predicted object in the to-be-predicted object set, multi-dimensional relationship data between the current to-be-predicted object and each first object in the to-be-predicted object set except the current to-be-predicted object is determined; and a third prediction value of the current to-be-predicted object is determined based on the multi-dimensional relationship data;

[0058] For each to-be-predicted object in the to-be-predicted object set, a fourth prediction value of the current to-be-predicted object is determined based on a resource remaining peak value of the current to-be-predicted object in the second preset time period;

[0059] Based on the first prediction value, the second prediction value, the third prediction value and the fourth prediction value of each to-be-predicted object in the to-be-predicted object set, an object value of each to-be-predicted object is determined.

[0060] The object value prediction method, device, computer device, storage medium and computer program product can cluster the object features of each to-be-predicted object in the to-be-predicted object set to obtain a plurality of object groups, determine the first prediction value of each to-be-predicted object in the same object group based on the resource remaining peak value of each to-be-predicted object in the first preset time period, input the object features of each to-be-predicted object in the to-be-predicted object set into the regression model respectively, obtain the second prediction value of each to-be-predicted object in the to-be-predicted object set through the regression model, determine the third prediction value of the current to-be-predicted object based on the multi-dimensional relationship data between the current to-be-predicted object and each first object in the to-be-predicted object set except the current to-be-predicted object, determine the fourth prediction value of the current to-be-predicted object based on the resource remaining peak value of the current to-be-predicted object in the second preset time period, and determine the object value of each to-be-predicted object based on the first prediction value, the second prediction value, the third prediction value and the fourth prediction value of each to-be-predicted object in the to-be-predicted object set. The method of predicting the object value from multiple angles such as clustering, regression model, multi-dimensional relationship data and resource remaining value can avoid the deviation caused by predicting the object value from a single angle, has a more comprehensive prediction angle, and improves the accuracy of the prediction result. BRIEF DESCRIPTION OF DRAWINGS

[0061] Figure 1 An application environment diagram of the object value prediction method in an embodiment;

[0062] Figure 2 A flowchart of the object value prediction method in an embodiment;

[0063] Figure 3 A sub-flowchart of S204 in an embodiment;

[0064] Figure 4 A sub-flowchart of S302 in an embodiment;

[0065] Figure 5 a sub-process flowchart of S308 in one embodiment;

[0066] Figure 6 a flowchart of an object value prediction method in another embodiment;

[0067] Figure 7 a structure block diagram of an object value prediction device in one embodiment;

[0068] Figure 8 an internal structure diagram of a computer device in one embodiment. DETAILED DESCRIPTION

[0069] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not used to limit the present application.

[0070] The object value prediction method provided by the embodiments of the present application can be applied to, for example, Figure 1The application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data required by the server 104 to process. The data storage system can be integrated on the server 104, or placed on the cloud or other network servers. The terminal 102 obtains the object characteristics of each to-be-predicted object in the to-be-predicted object set; by clustering the object characteristics of each to-be-predicted object, the to-be-predicted objects are divided into multiple object groups, and based on the resource remaining peak value of each to-be-predicted object in each object group within a first preset time period, the first prediction value of each to-be-predicted object in each object group is determined; the object characteristics of each to-be-predicted object in the to-be-predicted object set are input into the regression model respectively, and the second prediction value of each to-be-predicted object in the to-be-predicted object set is obtained; for each to-be-predicted object in the to-be-predicted object set, the multi-dimensional relationship data between the current to-be-predicted object and each first object in the to-be-predicted object set except the current to-be-predicted object is determined; based on the multi-dimensional relationship data, the third prediction value of the current to-be-predicted object is determined; for each to-be-predicted object in the to-be-predicted object set, based on the resource remaining peak value of the current to-be-predicted object within a second preset time period, the fourth prediction value of the current to-be-predicted object is determined; based on the first prediction value, the second prediction value, the third prediction value and the fourth prediction value of each to-be-predicted object in the to-be-predicted object set, the object value of each to-be-predicted object is determined. Among them, the terminal 102 can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things device can be a smart speaker, a smart TV, a smart air conditioner, a smart vehicle-mounted device, etc. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, etc. The server 104 can be realized by an independent server or a server cluster composed of multiple servers.

[0071] In one embodiment, as shown in Figure 2 , an object value prediction method is provided, which is applied to the terminal 102 in Figure 1 for example, including the following steps:

[0072] S201, obtaining the object characteristics of each to-be-predicted object in the to-be-predicted object set.

[0073] The set of to-be-predicted objects includes at least one to-be-predicted object, and the to-be-predicted object is an object that needs to be predicted in value. The object feature refers to a feature of the to-be-predicted object related to a specific business. The type of the object feature includes any one of a state feature, a product feature, or an event feature. For example, the state feature includes, but is not limited to, any one of basic information of the object, an account opening situation, or a business state. The product feature includes, but is not limited to, any one of a resource transfer situation, a resource holding amount, or an object level. The event feature includes, but is not limited to, any one of a resource transfer-in and transfer-out event, a resource transfer flow, or a login flow.

[0074] The terminal obtains the object feature of each to-be-predicted object in the set of to-be-predicted objects, including that the terminal obtains business data of each to-be-predicted object in the set of to-be-predicted objects, and performs feature engineering processing on the business data to obtain the object feature of each to-be-predicted object in the set of to-be-predicted objects. Feature engineering refers to filtering better data features from business data in a series of engineering ways to improve the training effect of the model. Feature engineering includes, but is not limited to, any one of data preprocessing, feature selection, and dimension reduction. For example, feature engineering also includes missing value supplement, which refers to replacing a null value of a numerical type in the business data with a preset numerical value, and replacing a null value or an empty string of a character type in the business data with a preset character. For example, feature engineering also includes one-hot encoding, which refers to replacing a category variable in the business data with multiple variables, introducing a new feature for each category, thereby representing any number of categories.

[0075] In S202, each to-be-predicted object is divided into a plurality of object groups by clustering the object features of each to-be-predicted object, and a first predicted value of each to-be-predicted object in each object group is determined based on a resource remaining peak value of each to-be-predicted object in the first preset time period in each object group.

[0076] The clustering refers to a method of classifying the object features of each to-be-predicted object by using a clustering algorithm. The clustering algorithm is based on similarity. Each object group obtained by the clustering algorithm has a high similarity between the to-be-predicted objects in each object group. Common clustering algorithms include, but are not limited to, Kmeans, DBSCAN (Density-Based Spatial Clustering of Applications with Noise), spectral clustering, Laplacian mapping, PCA (Principal Component Analysis), GMM (Gaussian Mixture Model), MeanShift, and hierarchical clustering. Specifically, the terminal classifies the to-be-predicted objects into multiple object groups by clustering the object features of each to-be-predicted object.

[0077] The resource residual peak value refers to the largest resource residual value in the resource residual value distribution of the to-be-predicted object within a preset time period. The object value prediction based on the clustering algorithm can improve the efficiency of the object value prediction. Each object group in the multiple object groups determined based on the clustering algorithm has a high similarity between the to-be-predicted objects in each object group. The terminal takes the resource residual peak value of each to-be-predicted object in each object group within the first preset time period as the first prediction value of each to-be-predicted object in each object group. The first prediction values of each to-be-predicted object in each object group are the same, which is conducive to obtaining the prediction values of each to-be-predicted object in each object group and improving the accuracy of the object value prediction.

[0078] S203, respectively input the object features of each to-be-predicted object in the set of to-be-predicted objects into the regression model to obtain the second prediction values of each to-be-predicted object in the set of to-be-predicted objects.

[0079] The regression model is a supervised learning machine learning model, which is used to establish a mapping relationship between independent variables and observation values. Common regression models include, but are not limited to, any one of linear regression, polynomial regression, ridge regression, lasso regression, elastic net regression, or XGBoost (eXtreme Gradient Boosting). The regression model is a trained machine learning model. Before the terminal inputs the object features of each to-be-predicted object in the set of to-be-predicted objects into the regression model, the terminal also needs to train the regression model until the prediction stopping condition is met. The trained machine learning model obtained is the regression model. The terminal inputs the object features of each to-be-predicted object in the set of to-be-predicted objects into the regression model to obtain the second prediction values of each to-be-predicted object in the set of to-be-predicted objects.

[0080] S204, for each to-be-predicted object in the to-be-predicted object set, determine multi-dimensional relationship data between the current to-be-predicted object and each first object in the to-be-predicted object set except the current to-be-predicted object; and determine a third prediction value of the current to-be-predicted object based on the multi-dimensional relationship data.

[0081] The first object refers to a to-be-predicted object in the to-be-predicted object set except the current to-be-predicted object. The multi-dimensional relationship data refers to relationship data between the to-be-predicted object and the first object in multiple dimensions. The multi-dimensional relationship data includes, but is not limited to, any one of resource transfer relationship data, resource recommendation relationship data, IP address intimacy, login address latitude and longitude similarity, or resource transfer organization relationship data. The multi-dimensional relationship between the current to-be-predicted object and the first object is directional, specifically, it can be one-way or two-way. For each to-be-predicted object in the to-be-predicted object set, the terminal determines the multi-dimensional relationship data between the current to-be-predicted object and each first object in the to-be-predicted object set except the current to-be-predicted object, and determines the third prediction value of the current to-be-predicted object according to the multi-dimensional relationship data between the current to-be-predicted object and the first object and the knowledge graph of the to-be-predicted object set. The knowledge graph of the to-be-predicted object set is a network knowledge structure reflecting the relationships of each to-be-predicted object in the to-be-predicted object set.

[0082] S205, for each to-be-predicted object in the to-be-predicted object set, determine a fourth prediction value of the current to-be-predicted object based on a resource remaining peak value of the current to-be-predicted object in a second preset time period.

[0083] For each to-be-predicted object in the to-be-predicted object set, the terminal obtains the resource remaining peak value of the current to-be-predicted object in the second preset time period, and takes the resource remaining peak value of the current to-be-predicted object in the second preset time period as the fourth prediction value of the current to-be-predicted object.

[0084] S206, determine an object value of each to-be-predicted object based on the first prediction value, the second prediction value, the third prediction value, and the fourth prediction value of each to-be-predicted object in the to-be-predicted object set.

[0085] The terminal fuses the first prediction value, the second prediction value, the third prediction value, and the fourth prediction value of each to-be-predicted object in the to-be-predicted object set, and takes the fusion result as the object value of each to-be-predicted object. The terminal fuses the prediction value results obtained by the four methods together, and the determined object value can satisfy the prediction conditions from four angles, and the object value has high accuracy.

[0086] In the object value prediction method, the object features of each to-be-predicted object in the to-be-predicted object set are clustered to obtain a plurality of object groups, the first prediction values of each to-be-predicted object in the same object group are determined based on the resource residual peak values of each to-be-predicted object in the first preset time period in each object group; the object features of each to-be-predicted object in the to-be-predicted object set are input into the regression model respectively, and the second prediction values of each to-be-predicted object in the to-be-predicted object set are obtained through the regression model; for each to-be-predicted object in the to-be-predicted object set, the third prediction value of the current to-be-predicted object is determined based on the multi-dimensional relationship data between the current to-be-predicted object and each first object in the to-be-predicted object set except the current to-be-predicted object; the fourth prediction value of the current to-be-predicted object is determined based on the resource residual peak value of the current to-be-predicted object in the second preset time period; and the object value of each to-be-predicted object is determined based on the first prediction value, the second prediction value, the third prediction value and the fourth prediction value of each to-be-predicted object in the to-be-predicted object set. This method of predicting the object value from multiple angles such as clustering, regression model, multi-dimensional relationship data and resource residual value can avoid the deviation caused by single-angle prediction of the object value, has a more comprehensive prediction angle, and improves the accuracy of the prediction result.

[0087] In one embodiment, the training step of the regression model includes: obtaining object features of a training object and a value label of the training object; inputting the object features of the training object into a current training model to obtain a predicted value of the training object, determining whether the current training model meets a preset stop condition based on the predicted value and the value label, if the current training model meets the preset stop condition, taking the current training model as the regression model, and if the current training model does not meet the preset stop condition, continuing to train the current training model until the current training model meets the preset stop condition.

[0088] The training object is an object whose object value is known. The value label is label information used to represent the object value of the training object. The current training model refers to the regression model before model training, and the current training model becomes the regression model after model training.

[0089] The terminal obtains the object feature of the training object and the value label of the training object, inputs the object feature of the training object into the current training model to obtain the predicted value of the training object. The terminal compares the predicted value of the training object with the value label to obtain a loss function of the current training model. It is determined whether the loss function is within a preset loss range. If the loss function is within the preset loss range, the current training model meets the preset condition, and the terminal takes the current training model as the regression model. If the loss function is not within the preset loss range, the current training model does not meet the preset condition, and the terminal continues to train the current training model until the loss function is within the preset range, and the current training model meets the preset stop condition. The loss function includes but is not limited to any one of RMSE (Root Mean Squared Errors), MSE (Mean Square Error), MAE (Mean Absolute Error), SD (Standard Deviation) or R2 (R-Square).

[0090] In the embodiment, the object feature of the training object and the value label of the training object are obtained, the object feature of the training object is input into the current training model to obtain the predicted value of the training object, and it is determined whether the current training model meets the preset stop condition based on the predicted value and the value label. The model training based on the value label of the training object can ensure that the regression model obtained by training has high object value prediction accuracy.

[0091] In one embodiment, as shown in FIG. 1, based on multi-dimensional relationship data, a third predicted value of a current to-be-predicted object is determined, including: Figure 3

[0092] S302, determining a sub-graph corresponding to the current to-be-predicted object on a knowledge graph of a to-be-predicted object set.

[0093] The knowledge graph is a knowledge structure in symbolic form describing resource transfer nodes and the order of resource transfer data flowing through the resource transfer nodes. The knowledge graph of the to-be-predicted object set includes each to-be-predicted object in the to-be-predicted object set. The sub-graph is a graph composed of nodes corresponding to the current to-be-predicted object and nodes directly or indirectly connected to the current to-be-predicted object on the knowledge graph. The terminal determines the sub-graph corresponding to the current to-be-predicted object on the knowledge graph of the to-be-predicted object set.

[0094] S304, for each first object in the to-be-predicted object set except the current to-be-predicted object, the multi-dimensional relationship data of the current to-be-predicted object and the current first object is weighted and averaged to obtain the close distance between the current to-be-predicted object and the current first object.​

[0095] wherein, for each first object in the set of objects to be predicted except the current object to be predicted, the terminal performs a weighted average of the multi-dimensional relationship data of the current object to be predicted and the current first object to obtain an affinity distance between the current object to be predicted and the current first object. In some embodiments, the multi-dimensional relationship data can be an affinity in the multi-dimensional, and the affinity ranges from 0 to 1. The closer the affinity is to 0, the more intimate the two objects are in the current dimensional relationship. The closer the affinity is to 1, the more distant the two objects are in the current dimensional relationship. Illustratively, the affinity between the current object to be predicted and the current first object in the dimension k is defined as wherein The affinity distance between the current object to be predicted and the current first object can be represented as:

[0096]

[0097] S306, obtaining a propagation weight between the current object to be predicted and the current first object according to the affinity distance between the current object to be predicted and the current first object.

[0098] wherein, the propagation weight refers to the weight of resource transfer between the current object to be predicted and the current first object, the greater the affinity distance, the smaller the propagation weight, the smaller the affinity distance, the greater the propagation weight. The terminal takes the difference between the affinity distance between the current object to be predicted and the current first object and 1 as the propagation weight between the current object to be predicted and the current first object.

[0099] S308, determining a third prediction value of the current object to be predicted in the sub-graph corresponding to the current object to be predicted based on the propagation weight between the current object to be predicted and each first object.

[0100] wherein, the terminal obtains the propagation weight between the current object to be predicted and each first object. In the sub-graph corresponding to the current object to be predicted, each first object only includes objects directly or indirectly related to the current object to be predicted. In the sub-graph corresponding to the current object to be predicted, the third prediction value of the current object to be predicted is determined in combination with the propagation weight between the current object to be predicted and the objects directly or indirectly related to the current object to be predicted.

[0101] In the embodiment, by determining the sub-knowledge graph corresponding to the current to-be-predicted object on the knowledge graph of the to-be-predicted object set, for each first object in the to-be-predicted object set, the multi-dimensional relationship data of the current to-be-predicted object and the current first object is weighted and averaged to obtain the close distance between the current to-be-predicted object and the current first object and the propagation weight, which can determine the third prediction value in the sub-knowledge graph of the current to-be-predicted object based on the multi-dimensional data between the current to-be-predicted object and each first object, and is beneficial to improve the object value prediction accuracy of the current to-be-predicted object.

[0102] In one embodiment, as shown in FIG. 1, Figure 4 determining the sub-knowledge graph corresponding to the current to-be-predicted object on the knowledge graph of the to-be-predicted object set includes:

[0103] S402, constructing a knowledge graph of a to-be-predicted object set, wherein each to-be-predicted object in the to-be-predicted object set corresponds to a node in the knowledge graph.

[0104] The terminal constructs the knowledge graph of the to-be-predicted object set. The basic component unit of the knowledge graph is an entity-relation-entity triple and an entity and its related attributes. Entities are connected to each other through relationships to form a knowledge graph. Each to-be-predicted object in the to-be-predicted object set corresponds to a node in the knowledge graph of the to-be-predicted object set. The relationship between each to-be-predicted object constitutes the relationship between entities in the knowledge graph of the to-be-predicted object set.

[0105] S404, selecting relationship nodes within a preset order relationship range of the current to-be-predicted object in the knowledge graph with the node corresponding to the current to-be-predicted object as the center.

[0106] The relationship nodes are directly or indirectly connected to the to-be-predicted object. If the current to-be-predicted object is directly connected to the relationship node, the relationship node is referred to as a relationship node within the first order relationship range of the to-be-predicted object. If the current to-be-predicted object is indirectly connected to the relationship node through a node, the relationship node is referred to as a relationship node within the second order relationship range of the to-be-predicted object. Correspondingly, if the current to-be-predicted object is indirectly connected to the relationship node through N nodes, the relationship node is referred to as a relationship node within the Nth order relationship range of the to-be-predicted object. The terminal selects the relationship nodes within the preset order relationship range of the current to-be-predicted object in the knowledge graph with the node corresponding to the current to-be-predicted object as the center.

[0107] S406, determining the sub-knowledge graph corresponding to the current to-be-predicted object based on the node corresponding to the current to-be-predicted object and the relationship node.

[0108] The terminal determines the node corresponding to the current to-be-predicted object and the relationship node as entities in a sub-graph corresponding to the current to-be-predicted object, and determines the connection relationship between the node corresponding to the current to-be-predicted object and the relationship node as a relationship between the entities in the sub-graph, thereby determining the sub-graph corresponding to the current to-be-predicted object.

[0109] In this embodiment, the knowledge graph of the to-be-predicted object set is constructed, the node corresponding to the current to-be-predicted object is taken as the center, and the relationship node within the preset order relationship range of the current to-be-predicted object is selected in the knowledge graph, thereby determining the sub-graph corresponding to the current to-be-predicted object. The sub-graph corresponding to the current to-be-predicted object is determined by the node corresponding to the current to-be-predicted object, the relationship node, and the connection relationship between the node corresponding to the current to-be-predicted object and the relationship node. The value of the current to-be-predicted object is predicted in the sub-graph corresponding to the current to-be-predicted object, which can improve the accuracy of the object value prediction.

[0110] In one embodiment, as shown in Figure 5 The third prediction value of the current to-be-predicted object is determined in the sub-graph corresponding to the current to-be-predicted object based on the propagation weight between the current to-be-predicted object and each first object, including:

[0111] S501, in the sub-graph corresponding to the current to-be-predicted object, determining a shortest path from each relationship node to the node corresponding to the current to-be-predicted object, and a target node passed through by the shortest path.

[0112] In the sub-graph corresponding to the current to-be-predicted object, each relationship node can directly or indirectly reach the node corresponding to the current to-be-predicted object. In the sub-graph corresponding to the current to-be-predicted object, the terminal determines a shortest path from each relationship node to the node corresponding to the current to-be-predicted object, and a target node passed through by the shortest path. In some embodiments, there is only one path from each relationship node to the node corresponding to the current to-be-predicted object, and the terminal determines the only path and the target node passed through by the only path. In other embodiments, there are multiple paths from each relationship node to the node corresponding to the current to-be-predicted object, and the terminal selects a shortest path from the multiple paths as the shortest path from each relationship node to the node corresponding to the current to-be-predicted object, and obtains the target node passed through by the shortest path.

[0113] S502, obtaining a propagation weight between two adjacent target nodes in the target nodes.

[0114] The terminal obtains the propagation weight between two adjacent target nodes in the target nodes. That is, the terminal obtains the propagation weight between the to-be-predicted objects corresponding to the two adjacent target nodes in the target nodes.

[0115] S503, obtaining a resource remaining peak value of the to-be-predicted object corresponding to each relationship node in a third preset time period.

[0116] The terminal obtains the resource remaining peak value of the to-be-predicted object corresponding to each relationship node in the third preset time period.

[0117] S504, determining an influence of the to-be-predicted object corresponding to each relationship node on the current to-be-predicted object based on the propagation weight between adjacent two target nodes in the target node and the resource remaining peak value of the to-be-predicted object corresponding to each relationship node in the third preset time period.

[0118] The terminal multiplies the propagation weight between adjacent two target nodes in the target node to determine a multiplication result, multiplies the multiplication result with the resource remaining peak value of the to-be-predicted object corresponding to each relationship node in the third preset time period to determine the influence of the to-be-predicted object corresponding to each relationship node on the current to-be-predicted object.

[0119] S505, summing the influences of the to-be-predicted objects corresponding to each relationship node on the current to-be-predicted object to determine a third prediction value of the current to-be-predicted object.

[0120] The terminal sums the influences of the to-be-predicted objects corresponding to each relationship node on the current to-be-predicted object to obtain a summation result, and takes the summation result as the third prediction value of the current to-be-predicted object. In some embodiments, P i is defined as the node corresponding to the current to-be-predicted object, each relationship node p j corresponds to a to-be-predicted object, and the resource remaining peak value of the to-be-predicted object in a third preset time period is α j , p j is a node on a shortest path from P i to P l , and the influence of node p j on node P i is:

[0121]

[0122] wherein k0=j, k l+1 =i, and the third prediction value V of node P i may be expressed as:

[0123] In this embodiment, the shortest path from each relationship node to the node corresponding to the current to-be-predicted object in the sub-graph corresponding to the current to-be-predicted object is determined, and the target nodes passed by the shortest path are determined. Based on the propagation weight and the resource residual peak value, the value influence corresponding to each relationship node is determined, and the third predicted value of the current to-be-predicted object is the sum of the value influences of each relationship node on the current to-be-predicted object. This method of determining the third predicted value by summing the value influences of each relationship node on the current to-be-predicted object in the sub-graph corresponding to the current to-be-predicted object can improve the accuracy of object value prediction.

[0124] In one embodiment, based on the first predicted value, the second predicted value, the third predicted value, and the fourth predicted value of each to-be-predicted object in the to-be-predicted object set, the object value of each to-be-predicted object is determined, including: for each to-be-predicted object in the to-be-predicted object set, selecting the maximum value from the first predicted value, the second predicted value, the third predicted value, and the fourth predicted value of the current to-be-predicted object, and taking the maximum value as the object value of the current to-be-predicted object.

[0125] In one embodiment, based on the first predicted value, the second predicted value, the third predicted value, and the fourth predicted value of each to-be-predicted object in the to-be-predicted object set, the object value of each to-be-predicted object is determined, including: for each to-be-predicted object in the to-be-predicted object set, selecting the maximum value from the first predicted value, the second predicted value, the third predicted value, and the fourth predicted value of the current to-be-predicted object, and taking the maximum value as the object value of the current to-be-predicted object.

[0126] In this embodiment, by taking the maximum value of the first predicted value, the second predicted value, the third predicted value, and the fourth predicted value of the current to-be-predicted object as the object value of the current to-be-predicted object, the deviation caused by single-angle prediction of the object value can be avoided, the prediction angle is more comprehensive, and the accuracy of the prediction result is improved.

[0127] To illustrate the object value prediction method and its effects in this scheme in detail, a most detailed embodiment is described as follows:

[0128] For the application field of object value prediction in the financial field. As shown in FIG. 1, the object value prediction method includes the following steps. Figure 6 As shown in FIG. 1, a flowchart of an object value prediction method is shown. A terminal obtains business data of each to-be-predicted object in a to-be-predicted object set, and performs feature engineering processing on the business data to obtain object features of each to-be-predicted object in the to-be-predicted object set.

[0129] The object features of each to-be-predicted object are clustered to divide the to-be-predicted objects into a plurality of object groups, and a first prediction value of each to-be-predicted object in each object group is determined based on a resource remaining peak value of each to-be-predicted object in the first object group within a first preset time period. The clustering algorithm used includes but is not limited to Mini-batch Kmeans (Mini-batch Kmeans algorithm). The clustering algorithm is evaluated using an MSE loss function.

[0130] The object features of each to-be-predicted object in the to-be-predicted object set are input into a regression model to obtain a second prediction value of each to-be-predicted object in the to-be-predicted object set. The regression model used includes but is not limited to an Xgboost regression model. The training steps of the regression model include: obtaining object features of a training object and a value label of the training object; inputting the object features of the training object into a current training model to obtain a predicted value of the training object, determining whether the current training model meets a preset stopping condition based on the predicted value and the value label, if yes, taking the current training model as the regression model, and if not, continuing to train the current training model until the current training model meets the preset stopping condition. The training object is an object whose object value is known. The value label is label information used to represent the object value of the training object. The current training model refers to the regression model before model training, and the regression model is obtained by training the current training model.

[0131] For each to-be-predicted object in the to-be-predicted object set, multi-dimensional relationship data between the current to-be-predicted object and each first object in the to-be-predicted object set except the current to-be-predicted object is determined, and a third prediction value of the current to-be-predicted object is determined based on the multi-dimensional relationship data. The multi-dimensional relationship data refers to the relationship data between the to-be-predicted object and the first object in multiple dimensions. The multi-dimensional relationship data includes but is not limited to any one of resource transfer relationship data, resource recommendation relationship data, IP address intimacy, similarity of login address latitude and longitude, or resource transfer organization relationship data.

[0132] Based on the multi-dimensional relationship data, the third prediction value of the current to-be-predicted object is determined, including: constructing a knowledge graph of the to-be-predicted object set, wherein each to-be-predicted object in the to-be-predicted object set corresponds to a node in the knowledge graph, selecting relationship nodes within a preset level relationship range of the current to-be-predicted object in the knowledge graph with the node corresponding to the current to-be-predicted object as the center, determining a sub-graph corresponding to the current to-be-predicted object based on the node corresponding to the current to-be-predicted object and the relationship nodes, and the sub-graph which can be expressed as:

[0133]

[0134] wherein, the node P iThe node corresponding to the current to-be-predicted object is denoted as node P j The relationship node is denoted as mind ji The distance of the shortest path from node P j to node P i is denoted as d i The m-order neighbor subgraph of node P ij is the subgraph spectrum corresponding to the current to-be-predicted object.

[0135] For each first object in the set of to-be-predicted objects except the current to-be-predicted object, the multi-dimensional relationship data of the current to-be-predicted object and the current first object is weighted and averaged to obtain the affinity distance between the current to-be-predicted object and the current first object. The multi-dimensional relationship data can be the affinity in the multi-dimensional space, and the affinity ranges from 0 to 1. The closer the affinity is to 0, the more intimate the two objects are in the current dimensional relationship. The closer the affinity is to 1, the more distant the two objects are in the current dimensional relationship. For example, the affinity between the current to-be-predicted object and the current first object in dimension k is defined as wherein The affinity distance between the current to-be-predicted object and the current first object can be represented as:

[0136]

[0137] According to the affinity distance between the current to-be-predicted object and the current first object, the propagation weight ω ij between the current to-be-predicted object and the current first object is obtained. ij = 1-d ij .

[0138] In the subgraph spectrum corresponding to the current to-be-predicted object, the shortest path from each relationship node to the node corresponding to the current to-be-predicted object is determined, as well as the target nodes passed through by the shortest path. The propagation weight between adjacent target nodes in the target nodes is obtained, and the resource residual peak value of the to-be-predicted object corresponding to each relationship node in the third preset time period is obtained. Based on the propagation weight between adjacent target nodes in the target nodes and the resource residual peak value of the to-be-predicted object corresponding to each relationship node in the third preset time period, the value influence of the to-be-predicted object corresponding to each relationship node on the current to-be-predicted object is determined. The value influences of the to-be-predicted objects corresponding to the relationship nodes on the current to-be-predicted object are summed up to determine the third predicted value of the current to-be-predicted object. The third predicted value of the current to-be-predicted object is defined as P i The node corresponding to the current to-be-predicted object is denoted as node P j The resource residual peak value of the to-be-predicted object corresponding to each relationship node p j in the third preset time period is denoted as a j The shortest path from node p i to node P l is denoted as j-k1-k2-…-k jThe value impact on node P i is:

[0139]

[0140] where k0=j, k l+1 =i, the third predicted value V of node P i may be expressed as:

[0141]

[0142] For each to-be-predicted object in the to-be-predicted object set, the terminal obtains the resource remaining peak value of the current to-be-predicted object in the second preset time period, and takes the resource remaining peak value of the current to-be-predicted object in the second preset time period as the fourth predicted value of the current to-be-predicted object.

[0143] For each to-be-predicted object in the to-be-predicted object set, the terminal selects the maximum value from the first predicted value, the second predicted value, the third predicted value, and the fourth predicted value of the current to-be-predicted object, and takes the maximum value as the object value of the current to-be-predicted object. In some embodiments, the terminal can also select any value from the first predicted value, the second predicted value, the third predicted value, and the fourth predicted value of the current to-be-predicted object as the object value of the current to-be-predicted object. In other embodiments, the terminal can also take the mean value of any number of values from the first predicted value, the second predicted value, the third predicted value, and the fourth predicted value of the current to-be-predicted object as the object value of the current to-be-predicted object.

[0144] The object value prediction method can obtain a plurality of object groups by clustering object features of each to-be-predicted object in the to-be-predicted object set, determine a first prediction value of each to-be-predicted object in the same object group based on resource residual peak values of each to-be-predicted object in each object group within a first preset time period, input the object features of each to-be-predicted object in the to-be-predicted object set into a regression model respectively, obtain second prediction values of each to-be-predicted object in the to-be-predicted object set through the regression model, determine a third prediction value of a current to-be-predicted object based on multi-dimensional relationship data between the current to-be-predicted object and each first object in the to-be-predicted object set except the current to-be-predicted object, determine a fourth prediction value of the current to-be-predicted object based on a resource residual peak value of the current to-be-predicted object within a second preset time period, and determine an object value of each to-be-predicted object based on the first prediction value, the second prediction value, the third prediction value, and the fourth prediction value of each to-be-predicted object in the to-be-predicted object set. The object value prediction method can avoid the deviation caused by predicting the object value from a single angle, has a more comprehensive prediction angle, and improves the accuracy of the prediction result.

[0145] It should be understood that, although each step in the flowchart involved in each embodiment as described above is shown in sequence according to the arrow, these steps are not necessarily executed in sequence according to the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other sequences. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least part of other steps or stages in other steps.

[0146] Based on the same inventive concept, the embodiments of the present application also provide an object value prediction device for implementing the object value prediction method described above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more object value prediction device embodiments provided below can refer to the limitations of the object value prediction method described above, which will not be repeated here.

[0147] In one embodiment, as Figure 7As shown, an object value prediction apparatus 100 is provided, comprising a feature acquisition module 110, a first value acquisition module 120, a second value acquisition module 130, a third value acquisition module 140, a fourth value acquisition module 150, and a value determination module 160, wherein:

[0148] The feature acquisition module 110 is configured to acquire object features of each to-be-predicted object in the to-be-predicted object set.

[0149] The first value acquisition module 120 is configured to cluster each to-be-predicted object in the to-be-predicted object set based on the object features of each to-be-predicted object, divide each to-be-predicted object into a plurality of object groups, and determine a first predicted value of each to-be-predicted object in each object group based on a resource remaining peak value of each to-be-predicted object in the first preset time period.

[0150] The second value acquisition module 130 is configured to input the object features of each to-be-predicted object in the to-be-predicted object set into a regression model respectively, and obtain a second predicted value of each to-be-predicted object in the to-be-predicted object set.

[0151] The third value acquisition module 140 is configured to determine, for each to-be-predicted object in the to-be-predicted object set, multi-dimensional relationship data between the current to-be-predicted object and each first object in the to-be-predicted object set except the current to-be-predicted object, and determine a third predicted value of the current to-be-predicted object based on the multi-dimensional relationship data.

[0152] The fourth value acquisition module 150 is configured to determine, for each to-be-predicted object in the to-be-predicted object set, a fourth predicted value of the current to-be-predicted object based on a resource remaining peak value of the current to-be-predicted object in the second preset time period.

[0153] The value determination module 160 is configured to determine an object value of each to-be-predicted object in the to-be-predicted object set based on the first predicted value, the second predicted value, the third predicted value, and the fourth predicted value of each to-be-predicted object.

[0154] The object value prediction device can obtain a plurality of object groups by clustering object features of each to-be-predicted object in the to-be-predicted object set, determine a first prediction value of each to-be-predicted object in the same object group based on a resource remaining peak value of each to-be-predicted object in each object group within a first preset time period, input the object features of each to-be-predicted object in the to-be-predicted object set into a regression model, obtain a second prediction value of each to-be-predicted object in the to-be-predicted object set through the regression model, determine a third prediction value of a current to-be-predicted object based on multi-dimensional relationship data between the current to-be-predicted object and each first object in the to-be-predicted object set except the current to-be-predicted object, determine a fourth prediction value of the current to-be-predicted object based on a resource remaining peak value of the current to-be-predicted object within a second preset time period, and determine an object value of each to-be-predicted object based on the first prediction value, the second prediction value, the third prediction value, and the fourth prediction value of each to-be-predicted object in the to-be-predicted object set. This method of predicting an object value from multiple angles such as clustering, a regression model, multi-dimensional relationship data, and a resource remaining value can avoid the deviation caused by predicting an object value from a single angle, is more comprehensive in the prediction angle, and improves the accuracy of the prediction result.

[0155] In one embodiment, in terms of training of the regression model, the second value obtaining module 130 is further configured to: obtain object features of a training object and a value label of the training object; input the object features of the training object into a current training model to obtain a predicted value of the training object, determine whether the current training model meets a preset stop condition based on the predicted value and the value label, if yes, take the current training model as the regression model, and if not, continue to train the current training model until the current training model meets the preset stop condition.

[0156] In one embodiment, in terms of determining the third prediction value of the current to-be-predicted object based on the multi-dimensional relationship data, the third value obtaining module 140 is further configured to: determine a sub-graph corresponding to the current to-be-predicted object on a knowledge graph of the to-be-predicted object set; for each first object in the to-be-predicted object set except the current to-be-predicted object, perform weighted average on the multi-dimensional relationship data between the current to-be-predicted object and the current first object to obtain an intimate distance between the current to-be-predicted object and the current first object; obtain a propagation weight between the current to-be-predicted object and the current first object according to the intimate distance between the current to-be-predicted object and the current first object; and determine the third prediction value of the current to-be-predicted object in the sub-graph corresponding to the current to-be-predicted object based on the propagation weight between the current to-be-predicted object and each first object.

[0157] In an embodiment, in determining the sub-knowledge graph corresponding to the current to-be-predicted object on the knowledge graph of the set of to-be-predicted objects, the third value obtaining module 140 is further configured to: construct the knowledge graph of the set of to-be-predicted objects, where each to-be-predicted object in the set of to-be-predicted objects corresponds to a node in the knowledge graph; select, in the knowledge graph, relationship nodes within a preset order of relationship range of the current to-be-predicted object, with the node corresponding to the current to-be-predicted object as the center; and determine the sub-knowledge graph corresponding to the current to-be-predicted object based on the node corresponding to the current to-be-predicted object and the relationship nodes.

[0158] In an embodiment, in determining the third predicted value of the current to-be-predicted object in the sub-knowledge graph corresponding to the current to-be-predicted object based on the propagation weight between the current to-be-predicted object and each first object, the third value obtaining module 140 is further configured to: determine, in the sub-knowledge graph corresponding to the current to-be-predicted object, a shortest path from each relationship node to the node corresponding to the current to-be-predicted object and a target node passed through by the shortest path; obtain the propagation weight between adjacent two target nodes in the target nodes; obtain a resource residual peak value of the to-be-predicted object corresponding to each relationship node within a third preset time period; determine an influence of the to-be-predicted object corresponding to each relationship node on the value of the current to-be-predicted object based on the propagation weight between adjacent two target nodes in the target nodes and the resource residual peak value of the to-be-predicted object corresponding to each relationship node within the third preset time period; and sum up the influences of the to-be-predicted objects corresponding to the relationship nodes on the value of the current to-be-predicted object to determine the third predicted value of the current to-be-predicted object.

[0159] In an embodiment, in determining the object value of each to-be-predicted object based on the first predicted value, the second predicted value, the third predicted value and the fourth predicted value of each to-be-predicted object in the set of to-be-predicted objects, the value determining module 160 is further configured to: for each to-be-predicted object in the set of to-be-predicted objects, select a maximum value from the first predicted value, the second predicted value, the third predicted value and the fourth predicted value of the current to-be-predicted object, and take the maximum value as the object value of the current to-be-predicted object.

[0160] The modules in the above object value prediction apparatus can be realized by software, hardware or a combination thereof. The modules can be embedded in or independent of a processor in a computer device in hardware form, or stored in a memory in the computer device in software form, so as to be called and executed by a processor to perform the operations corresponding to the modules.

[0161] In an embodiment, a computer device is provided, which can be a server, and an internal structure diagram of the computer device can be as shown in FIG. 8. Figure 8As shown in the figure. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through the system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capability. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the object set to be predicted, the object feature, the object group, the resource residual peak value, the regression model, the multi-dimensional relationship data, the first prediction value, the second prediction value, the third prediction value and the fourth prediction value. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with the terminal outside through the network connection. The computer program is executed by the processor to realize an object value prediction method.

[0162] Those skilled in the art can understand that, Figure 8 The skilled in the art can understand that,

[0163] In one embodiment, a computer device is provided, including a memory and a processor, and the memory stores a computer program, which is executed by the processor to realize the steps in the above method embodiments.

[0164] In one embodiment, a computer readable storage medium is provided, which stores a computer program, which is executed by the processor to realize the steps in the above method embodiments.

[0165] In one embodiment, a computer program product is provided, including a computer program, which is executed by the processor to realize the steps in the above method embodiments.

[0166] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0167] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (Phase Change Memory, PCM), graphene memory, etc. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0168] Any combination of the technical features of the above embodiments can be made. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0169] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of protection of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. An object value prediction method characterized by, The method comprises: acquiring object features of each to-be-predicted object in a to-be-predicted object set; grouping each to-be-predicted object into a plurality of object groups by clustering the object features of each to-be-predicted object, and determining a first prediction value of each to-be-predicted object in each object group based on a resource remaining peak value of each to-be-predicted object in the first preset time period; inputting the object features of each to-be-predicted object in the to-be-predicted object set into a regression model respectively to obtain a second prediction value of each to-be-predicted object in the to-be-predicted object set; for each to-be-predicted object in the to-be-predicted object set, determining multi-dimensional relationship data between the current to-be-predicted object and each first object except the current to-be-predicted object in the to-be-predicted object set, and determining a third prediction value of the current to-be-predicted object based on the multi-dimensional relationship data; for each to-be-predicted object in the to-be-predicted object set, determining a fourth prediction value of the current to-be-predicted object based on a resource remaining peak value of the current to-be-predicted object in a second preset time period; determining an object value of each to-be-predicted object based on the first prediction value, the second prediction value, the third prediction value and the fourth prediction value of each to-be-predicted object in the to-be-predicted object set.

2. The method of claim 1, wherein, The training step of the regression model comprises: acquiring object features of a training object and a value label of the training object; inputting the object features of the training object into a current training model to obtain a prediction value of the training object, determining whether the current training model meets a preset stop condition based on the prediction value and the value label, if yes, taking the current training model as the regression model, and if no, continuing to train the current training model until the current training model meets the preset stop condition.

3. The method of claim 1, wherein, The determination of the third prediction value of the current to-be-predicted object based on the multi-dimensional relationship data comprises: determining a sub-graph corresponding to the current to-be-predicted object on a knowledge graph of the to-be-predicted object set; for each first object except the current to-be-predicted object in the to-be-predicted object set, performing weighted average on the multi-dimensional relationship data between the current to-be-predicted object and the current first object to obtain an intimate distance between the current to-be-predicted object and the current first object; obtaining a propagation weight between the current to-be-predicted object and the current first object according to the intimate distance between the current to-be-predicted object and the current first object; determining the third prediction value of the current to-be-predicted object in the sub-graph corresponding to the current to-be-predicted object based on the propagation weight between the current to-be-predicted object and each first object.

4. The method of claim 3, wherein, The determination of the sub-graph corresponding to the current to-be-predicted object on the knowledge graph of the to-be-predicted object set comprises: constructing a knowledge graph of the to-be-predicted object set, wherein each to-be-predicted object in the to-be-predicted object set corresponds to a node in the knowledge graph; selecting relationship nodes within a preset level relationship range of the current to-be-predicted object in the knowledge graph with the node corresponding to the current to-be-predicted object as the center. Determine a sub-graph corresponding to the current to-be-predicted object based on the node corresponding to the current to-be-predicted object and the relationship node.

5. The method of claim 4, wherein, The third prediction value of the current to-be-predicted object is determined in the sub-graph corresponding to the current to-be-predicted object based on the propagation weight between the current to-be-predicted object and each first object, including: In the sub-graph corresponding to the current to-be-predicted object, the shortest path from each relationship node to the node corresponding to the current to-be-predicted object and the target node passed by the shortest path are determined. The propagation weight between adjacent two target nodes in the target node is obtained. The resource residual peak value of the to-be-predicted object corresponding to each relationship node in a third preset time period is obtained. The value influence of the to-be-predicted object corresponding to each relationship node on the current to-be-predicted object is determined based on the propagation weight between adjacent two target nodes in the target node and the resource residual peak value of the to-be-predicted object corresponding to each relationship node in the third preset time period. The value influence of the to-be-predicted object corresponding to each relationship node on the current to-be-predicted object is summed up to determine the third prediction value of the current to-be-predicted object.

6. The method of claim 1, wherein, The object value of each to-be-predicted object is determined based on the first prediction value, the second prediction value, the third prediction value and the fourth prediction value of each to-be-predicted object in the set of to-be-predicted objects, including: For each to-be-predicted object in the set of to-be-predicted objects, the maximum value is selected from the first prediction value, the second prediction value, the third prediction value and the fourth prediction value of the current to-be-predicted object, and the maximum value is taken as the object value of the current to-be-predicted object.

7. An object value prediction device characterized by comprising: The device comprises: A feature acquisition module is configured to obtain object features of each to-be-predicted object in a set of to-be-predicted objects. A first value acquisition module is configured to cluster the object features of each to-be-predicted object to divide the to-be-predicted objects into a plurality of object groups, and determine the first prediction value of each to-be-predicted object in each object group based on the resource residual peak value of each to-be-predicted object in the first object group in a first preset time period. A second value acquisition module is configured to input the object features of each to-be-predicted object in the set of to-be-predicted objects into a regression model respectively to obtain the second prediction value of each to-be-predicted object in the set of to-be-predicted objects. A third value acquisition module is configured to determine, for each to-be-predicted object in the set of to-be-predicted objects, multi-dimensional relationship data between the current to-be-predicted object and each first object except the current to-be-predicted object in the set of to-be-predicted objects, and determine the third prediction value of the current to-be-predicted object based on the multi-dimensional relationship data. A fourth value acquisition module is configured to determine, for each to-be-predicted object in the set of to-be-predicted objects, the fourth prediction value of the current to-be-predicted object based on the resource residual peak value of the current to-be-predicted object in a second preset time period. A value determination module is configured to determine the object value of each to-be-predicted object based on the first prediction value, the second prediction value, the third prediction value and the fourth prediction value of each to-be-predicted object in the set of to-be-predicted objects.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 6.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 6.

Images