An image data generation method, device, computer equipment and storage medium

Abstract

The application provides a portrait data generation method and device, computer equipment and a storage medium, comprising: identifying to-be-identified information in a database; extracting customer data in the to-be-identified information, and dividing the customer data into a target training set and a target test set; training a preset initial model through the target training set to obtain a mature model, testing the mature model through the target test set to obtain a test accuracy rate; if it is determined that the test accuracy rate is greater than a preset upper threshold, it is determined that the customer information is portrait data; if it is determined that the test accuracy rate is less than a preset lower threshold, it is determined that the content of the customer information in an update time period is portrait data, and invalid information is generated. The application ensures the reliability of generating user portraits, thereby ensuring the accuracy and timeliness of the finally obtained user portraits, enhancing the robustness of the intelligent recommendation system, and avoiding the occurrence of customer aversion.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to an image data generation method and device, a computer device and a storage medium. BACKGROUND

[0002] Customer information represents the attribute characteristics of a customer, such as work unit, address information, salary, and the like. In a bank intelligent recommendation system, customer information is important data for generating a user image. The intelligent recommendation system usually generates a user image based on customer information and recommends relevant products to the customer.

[0003] However, the current bank system often fails to update customer information in the intelligent recommendation system in a timely manner, resulting in inaccurate user images generated by the intelligent recommendation system and untimely problems.

[0004] To solve the above problems, the current bank system usually re-enters customer information into the intelligent recommendation system to generate a user image.

[0005] However, the inventors have found that customer information changes over time, and there can be a large difference between the attribute characteristics of a customer in the distant past and the attribute characteristics of the customer at present. If a user image is generated directly by the intelligent recommendation system based on customer information, the user image can be inaccurate, which can affect the robustness of the intelligent recommendation system. SUMMARY

[0006] The present application provides an image data generation method and device, a computer device and a storage medium to solve the problem that the current generation of a user image directly by an intelligent recommendation system based on customer information can result in inaccurate user images, which can affect the robustness of the intelligent recommendation system.

[0007] In a first aspect, the present application provides an image data generation method, comprising:

[0008] identifying to-be-identified information in a database; wherein the database stores at least one customer information; the to-be-identified information is customer information that has not been updated in content for more than a preset update time; and the customer information represents the attribute characteristics of a customer;

[0009] extracting customer data in the to-be-identified information, and dividing the customer data into a target training set and a target test set; wherein the customer data represents the transaction behavior and content of a customer in the past; the target training set is used to train a machine learning model; and the target test set is used to verify the trained machine learning model and predict the accuracy of the labels of the customer data in the target test set;

[0010] training the preset initial model by the target training set to obtain a mature model, and testing the mature model by the target test set to obtain a test accuracy rate; wherein the initial model comprises at least one machine learning model; the test accuracy rate represents the accuracy of the mature model in predicting the label of the customer data in the target test set; the label represents the collection time period in which the customer information is located; the collection time period comprises a historical time period and an update time period; the update time period is a time period from the last update time to the current time; the last update time represents the time when the content of the customer information was last updated; the historical time period is a time period in which the end point coincides with the start point of the update time period;

[0011] If it is determined that the test accuracy rate is greater than a preset upper threshold, it is determined that the customer information is portrait data; wherein the portrait data is output to an intelligent recommendation system; the intelligent recommendation system is used to generate a user portrait of a customer corresponding to the customer information according to the portrait data;

[0012] If it is determined that the test accuracy rate is less than a preset lower threshold, it is determined that the content of the customer information in the update time period is the portrait data, and invalidation information is generated; wherein the invalidation information represents the content of the customer information in the historical time period as invalid.

[0013] In the above scheme, the to-be-identified information in the database is identified, comprising:

[0014] Any customer information in the database is taken as target information, and the last update time of the target information is obtained from the database;

[0015] If it is determined that the time difference between the last update time and the current time exceeds the update time, it is determined that the target information is to-be-determined information;

[0016] If it is determined that the attribute information in the to-be-determined information changes, the to-be-determined information is set as to-be-identified information.

[0017] In the above scheme, if it is determined that the attribute information in the to-be-determined information changes, the to-be-determined information is set as to-be-identified information, comprising:

[0018] The attribute information of the target information in the collection time period is extracted; wherein the attribute information comprises first attribute information and second attribute information; the first attribute information is used to represent the attribute characteristics of the customer corresponding to the target information during the historical time period; the second attribute information is used to represent the attribute characteristics of the customer corresponding to the target information during the update time period;

[0019] If it is determined that the first attribute information and the second attribute information are inconsistent, the target information is taken as to-be-identified information.

[0020] In the above scheme, the customer data is divided into a target training set and a target test set, comprising:

[0021] Extract at least one customer data in the to-be-identified information, and divide at least one customer data into a first data set and a second data set; wherein the customer data represents the transaction behavior of the customer corresponding to the customer information in the preset collection time period; the collection time period includes the historical time period and the update time period, and the end point of the historical time period coincides with the start point of the update time; the customer data in the first data set represents the transaction behavior of the customer in the historical time period; the customer data in the second data set represents the transaction behavior of the customer in the update time period;

[0022] Label the customer data in the first data set with a first label, and label the customer data in the second data set with a second label; wherein the first label represents the customer information of the customer in the historical time period; the second label represents the customer information of the customer in the update time period;

[0023] Split the first data set into a first training set and a first test set, split the second data set into a second training set and a second test set, and aggregate the first training set and the second training set to obtain a target training set, and aggregate the first test set and the second test set to obtain a target test set.

[0024] In the above scheme, the target training set is used to train the preset initial model to obtain a mature model, comprising:

[0025] The first structured data in the target training set is used to train the random forest classification model in the initial model to obtain a mature forest classification model; wherein the first structured data refers to customer data with structured data structure; the structured data refers to data that can be represented and stored using a relational database, and / or can be logically expressed and implemented using a two-dimensional table;

[0026] The text vector of the second structured data in the target training set is calculated, and the text vector is used to train the logistic regression classification model in the initial model to obtain a mature regression classification model; wherein the second structured data refers to customer data with unstructured data structure; the unstructured data refers to data without fixed structure; the text vector is used to represent the word frequency and inverse document frequency of the unstructured customer data;

[0027] The mature forest classification model and / or the mature regression classification model are aggregated to obtain a mature model.

[0028] In the scheme, the test accuracy of the mature model is obtained by testing the mature model through the target test set, and the test accuracy comprises:

[0029] The first structure data in the target test set is labeled and predicted through a mature forest classification model in the mature model to obtain a forest verification label of the first structure data, wherein the forest verification label is a prediction result of the mature forest classification model in the label prediction of the first structure data.

[0030] The forest verification label of each first structure data is compared with the first label or the second label of the first structure data to determine a forest accuracy of the mature forest classification model:

[0031] The second structure data in the target test set is labeled and predicted through a mature regression classification model in the mature model to obtain a regression verification label of the second structure data, wherein the regression verification label is a prediction result of the mature regression classification model in the label prediction of the second structure data.

[0032] The regression verification label of each second structure data is compared with the first label or the second label of the second structure data to determine a regression accuracy of the regression classification model.

[0033] The test accuracy of the mature model is obtained according to the forest accuracy and the regression accuracy.

[0034] In the scheme, the test accuracy of the mature model is obtained according to the forest accuracy and the regression accuracy, and the test accuracy comprises:

[0035] An average value of the forest accuracy and the regression accuracy is calculated, and the average value is set as the test accuracy; or

[0036] A weighted value is obtained by calculating the forest accuracy and the regression accuracy through a preset weight function, and the weighted value is set as the test accuracy; or

[0037] An accuracy rate set is formed by summarizing the forest accuracy and the regression accuracy, and the accuracy rate set is set as the test accuracy.

[0038] In the scheme, before the to-be-identified information in the database is identified, the method further comprises:

[0039] The online transaction information and the offline transaction information of the target customer are collected.

[0040] The customer data corresponding to the structured label in the online transaction information and / or the offline transaction information is taken as the first structure data.

[0041] corresponding to the unstructured label in the offline transaction information as the second structured data;

[0042] corresponding to the unstructured label associated with the unstructured label in the offline transaction information as the second structured data.

[0043] In the above scheme, after the mature model is tested by the target test set to obtain a test accuracy, the method further comprises:

[0044] If it is determined that the test accuracy is less than or equal to the upper threshold value and greater than or equal to a preset lower threshold value, a weight value of content of the customer information in a historical time period is reduced to obtain first correction information, and / or a weight value of content of the customer information in an update time period is increased to obtain second correction information;

[0045] The first correction information and the second correction information are summarized, or the content of the customer information in the historical time period and the second correction information are summarized, or the first correction information and the content of the customer information in the update time period are summarized to obtain portrait data.

[0046] In the above scheme, reducing the weight value of the content of the customer information in the historical time period to obtain the first correction information comprises:

[0047] A first difference value between the test accuracy and the lower threshold value is calculated, and a first weight parameter corresponding to the first difference value is obtained through a preset first correction rule; wherein the first correction rule is used to represent a mapping relationship between different differences between the test accuracy and the lower threshold value and different weight parameters;

[0048] The first weight parameter is inserted into the content of the customer information in the historical time period to reduce the weight value of the content in the historical time period and obtain the first correction information.

[0049] In the above scheme, increasing the weight value of the content of the customer information in the update time period to obtain the second correction information comprises:

[0050] A second difference value between the test accuracy and the upper threshold value is calculated, and a second weight parameter corresponding to the second difference value is obtained through a preset second correction rule; wherein the second correction rule is used to represent a mapping relationship between different differences between the test accuracy and the upper threshold value and different weight parameters;

[0051] The second weight parameter is inserted into the content of the customer information in the update time period to increase the weight value of the content in the update time period and obtain the second correction information.

[0052] In a third aspect, the present application provides an image data generation device, comprising:

[0053] The acquisition module is configured to identify to-be-identified information in a database; the database stores at least one customer information; the to-be-identified information is customer information that has not been updated for more than a preset update time; and the customer information represents the attribute characteristics of a customer.

[0054] The preprocessing module is configured to extract customer data in the to-be-identified information, and divide the customer data into a target training set and a target test set; the customer data represents the transaction behavior and content of a customer in the past; the target training set is used to train a machine learning model; and the target test set is used to verify the trained machine learning model and predict the accuracy of the labels of the customer data in the target test set.

[0055] The training and prediction module is configured to train a preset initial model to obtain a mature model through the target training set, and test the mature model through the target test set to obtain a test accuracy; the initial model includes at least one machine learning model; the test accuracy represents the accuracy of the mature model in predicting the labels of the customer data in the target test set; the label represents the collection time period of the customer information; the collection time period includes a historical time period and an update time period; the update time period is a time period from the last update time to the current time; the last update time represents the time when the content of the customer information was last updated; and the historical time period is a time period in which the end point coincides with the start point of the update time period.

[0056] The first comparison and identification module is configured to determine that the customer information is image data if it is determined that the test accuracy is greater than a preset upper threshold; the image data is output to an intelligent recommendation system; and the intelligent recommendation system is configured to generate a user image of a customer corresponding to the customer information according to the image data.

[0057] The second comparison and identification module is configured to determine that the content of the customer information in the update time period is image data and generate invalid information if it is determined that the test accuracy is less than a preset lower threshold; and the invalid information represents the invalidity of the content of the customer information in the historical time period.

[0058] In a third aspect, the present application provides a computer device, comprising a processor and a memory in communication connection with the processor.

[0059] The memory stores computer execution instructions.

[0060] The processor executes the computer execution instructions stored in the memory to implement the above method.

[0061] In a fourth aspect, the present application provides a computer readable storage medium, wherein computer execution instructions are stored in the computer readable storage medium, and the computer execution instructions are used to implement the method described above when executed by a processor.

[0062] In a fifth aspect, the present application provides a computer program product, comprising a computer program, wherein the computer program is used to implement the method described above when executed by a processor.

[0063] The portrait data generation method, device, computer device and storage medium provided by the present application can identify the customer information in the database that is not updated in time by identifying the to-be-identified information in the database.

[0064] The customer data in the to-be-identified information is extracted, the customer data is divided into a target training set and a target test set, the initial model is trained by the target training set to obtain a mature model, and the mature model is tested by the target test set to obtain a test accuracy rate, so as to determine whether the machine learning model can accurately determine the collection time period of the customer data based on the customer data.

[0065] If the test accuracy rate is greater than the upper threshold, it indicates that the machine learning model trained by the customer data can accurately identify which time period of the customer information each customer data is in, and thus the content of the customer information in the historical time period and the content of the customer information in the update time period are explained, even if they are different, it will not affect the portrait label of the user portrait, so the customer information can be directly used as the portrait data for generating the user portrait.

[0066] If the test accuracy rate is less than the lower threshold, it indicates that the machine learning model trained by the customer data cannot accurately identify which time period of the customer information each customer data is in, and thus it is explained that the content of the customer information in the historical time period has been invalidated, and therefore the content of the customer information in the update time period needs to be used as the portrait data for generating the user portrait.

[0067] Therefore, by identifying the customer data based on the machine learning model and then determining the content of the customer information that can be used as the portrait data, the reliability for generating the user portrait is ensured, and thus the accuracy and timeliness of the obtained user portrait are ensured, the robustness of the intelligent recommendation system is enhanced, and the occurrence of customer dissatisfaction is avoided. BRIEF DESCRIPTION OF DRAWINGS

[0068] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0069] Figure 1 Structure diagram of information management system applied to the portrait data generation method of the present application;

[0070] Figure 2 Flow chart of embodiment 1 of the portrait data generation method of the present application;

[0071] Figure 3 Page recording first structure data and its label in embodiment 1 of the portrait data generation method of the present application;

[0072] Figure 4 Page recording second structure data and its label in embodiment 1 of the portrait data generation method of the present application;

[0073] Figure 5 Page of failure information generated in embodiment 1 of the portrait data generation method of the present application if the test accuracy rate is less than 0.3;

[0074] Figure 6 Flow chart of embodiment 2 of the portrait data generation method of the present application;

[0075] Figure 7 Page of searching the income and expenditure details of the user within two years in the TransDetails table according to the user ID and current time in embodiment 2 of the portrait data generation method of the present application;

[0076] Figure 8 Page of collecting the required merchant information through unstructured label and offline label if the transaction channel is offline transaction in embodiment 2 of the portrait data generation method of the present application;

[0077] Figure 9 Module diagram of embodiment 3 of the portrait data generation device of the present application;

[0078] Figure 10 Hardware structure diagram of the computer device in the computer device of the present application.

[0079] The specific embodiments of the present application have been shown in the above-described drawings, and will be described in more detail hereinafter. These drawings and textual descriptions are not intended to limit the scope of the concept of the present application by any means, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0080] The exemplary embodiments will be described in detail herein with reference to the attached drawings. The following description is made with reference to the accompanying drawings in which like reference numerals refer to like elements, unless the context of use indicates otherwise. The following description of exemplary embodiments is not representative of all embodiments consistent with the present application. Rather, it is merely an example of apparatus and methods consistent with some aspects of the present application as detailed in the appended claims.

[0081] 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, and provide corresponding operation portal for user to choose authorization or refusal.

[0082] It should be noted that the portrait data generation method, device, computer device and storage medium of the present application can be used in the field of artificial intelligence technology, and can also be used in any field other than the field of artificial intelligence technology. The application field of the portrait data generation method, device, computer device and storage medium of the present application is not limited.

[0083] The specific application scenario of the present application is:

[0084] Please refer to Figure 1 , a server 11 running a portrait data generation method is connected with a database 12 and an intelligent recommendation system 13;

[0085] The server 11 identifies the to-be-identified information in the database 12; wherein the database stores at least one customer information; the to-be-identified information is customer information that has not been updated in content for more than a preset update time; and the customer information represents the attribute characteristics of the customer;

[0086] The server 11 extracts customer data in the to-be-identified information, and divides the customer data into a target training set and a target test set; wherein the customer data represents the customer's historical transaction behavior and content; the target training set is used to train the machine learning model; and the target test set is used to verify the trained machine learning model, and predict the accuracy of the label of the customer data in the target test set;

[0087] The server 11 trains the preset initial model through the target training set to obtain a mature model, and tests the mature model through the target test set to obtain a test accuracy rate; wherein, the initial model comprises at least one machine learning model; the test accuracy rate represents the accuracy of the mature model in predicting the label of the customer data in the target test set; the label represents the collection time period of the customer information; the collection time period comprises a historical time period and an updated time period; the updated time period is a time period from the last update time to the current time; the last update time represents the time when the content of the customer information was last updated; the historical time period is a time period in which the end point coincides with the start point of the updated time period;

[0088] If the server 11 determines that the test accuracy rate is greater than the preset upper threshold, it is determined that the customer information is portrait data; wherein, the portrait data is output to the intelligent recommendation system 13; the intelligent recommendation system 13 is used to generate a user portrait of the customer corresponding to the customer information according to the portrait data;

[0089] If the server 11 determines that the test accuracy rate is less than the preset lower threshold, it is determined that the content of the customer information in the updated time period is the portrait data, and invalid information is generated; wherein, the invalid information represents that the content of the customer information in the historical time period is invalid.

[0090] By identifying the customer data based on the machine learning model, the content of the customer information that can be used as the portrait data is determined, which ensures the reliability of generating the user portrait, and thus ensures the accuracy and timeliness of the obtained user portrait, enhances the robustness of the intelligent recommendation system, and avoids the occurrence of customer complaints.

[0091] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems will be described in detail below with specific examples. The following specific examples can be combined with each other, and the same or similar concepts or processes may not be described again in some examples. The embodiments of the present application will be described below with reference to the accompanying drawings.

[0092] Embodiment 1:

[0093] Please refer to Figure 2 The present application provides a portrait data processing method, comprising:

[0094] S201: identifying the to-be-identified information in the database; wherein, the database stores at least one customer information; the to-be-identified information is customer information that has not been updated in content beyond a preset update time; the customer information represents the attribute characteristics of the customer.

[0095] In this step, the customer information that is not updated in time in the database is identified by identifying the to-be-identified information in the database.

[0096] In a preferred embodiment, the to-be-identified information in the database is identified, comprising:

[0097] Taking any customer information in the database as target information, obtaining the last update time of the target information from the database; wherein the last update time represents the time when the content of the customer information was last updated;

[0098] If it is determined that the time difference between the last update time and the current time exceeds the update time, it is determined that the target information is pending information;

[0099] If it is determined that the attribute information in the pending information has changed, the pending information is set as to-be-identified information.

[0100] For example, when the user logs in, at least one customer information is recorded in the STATUS field of the User table, which is used to mark customers who have not updated information for more than a year, and the status is "not updated". If it is determined that the target information is marked as "not updated", the marking module is called to receive the time of the last update information of the customer information, and it is determined whether it has exceeded one year from the current time. If it has exceeded, the STATUS field is updated, and the target information in the STATUS field is set as pending information.

[0101] Further, if it is determined that the attribute information in the pending information has changed, the pending information is set as to-be-identified information, comprising:

[0102] Extracting attribute information of the target information in a preset collection time period; wherein the collection time period includes a historical time period and an update time period; the end point of the historical time period coincides with the start point of the update time period; the update time period is a time period from the last update time to the current time; the attribute information includes first attribute information and second attribute information; the first attribute information is used to represent the attribute characteristics of the customer corresponding to the target information during the historical time period; the second attribute information is used to represent the attribute characteristics of the customer corresponding to the target information during the update time period;

[0103] If it is determined that the first attribute information and the second attribute information are inconsistent, the target information is set as to-be-identified information.

[0104] Optionally, the length of the update time period is consistent with the length of the historical time period, so as to ensure that the data quantity between the first data set and the second data set obtained subsequently can correspond, and ensure the reliability of the portrait data change degree identification.

[0105] For example, the update time period is 1 year, and the historical time period is 1 year, i.e. the historical time period is the first year, and the update time period is the second year; if the customer information (work unit, salary, address information, etc.) in the first year is different from the customer information in the second year, the target information, i.e. the different customer information, is set as to-be-identified information.

[0106] S202: Extract customer data from the information to be identified, and divide the customer data into a target training set and a target test set; wherein, the customer data represents the customer's historical transaction behavior and content; the target training set is used to train the machine learning model; the target test set is used to verify the trained machine learning model and predict the accuracy of the labels of the customer data in the target test set.

[0107] In a preferred embodiment, customer data is divided into a target training set and a target test set, including:

[0108] Extract at least one customer data from the information to be identified, and divide the at least one customer data into a first dataset and a second dataset; wherein, the customer data represents the transaction behavior of the customer corresponding to the customer information within a preset collection period; the collection period includes a historical period and an update period, the end of the historical period coincides with the start of the update period; the customer data in the first dataset represents the transaction behavior of the customer within the historical period; the customer data in the second dataset represents the transaction behavior of the customer within the update period;

[0109] The customer data in the first dataset is labeled with a first label, and the customer data in the second dataset is labeled with a second label; wherein, the first label represents the customer information in the historical time period; and the second label represents the customer information in the updated time period.

[0110] The first dataset is split into a first training set and a first test set. The second dataset is split into a second training set and a second test set. The first training set and the second training set are combined to obtain the target training set. The first test set and the second test set are combined to obtain the target test set.

[0111] For example, please refer to Figure 3 and Figure 4 Based on the example above, assume the obtained customer data consists of income and expense details, along with related merchant information. The first year's income and expense details are divided into two parts, one for training and the other for testing, with each set labeled as 1. Different detail options are extracted based on different customer information. The second year's income and expense details are processed in the same way, with each set labeled as 0.

[0112] in, Figure 3 The document contains the first structured data and its tags; the first structured data is customer data corresponding to structured tags in online transaction information and / or offline transaction information; Figure 4 The document contains second-structured data and its tags; the second-structured data is customer data in offline transaction information that corresponds to unstructured tags.

[0113] S203: training the preset initial model through the target training set to obtain a mature model, and testing the mature model through the target test set to obtain a test accuracy rate; wherein the initial model comprises at least one machine learning model; the test accuracy rate represents the accuracy of the mature model in predicting the label of the customer data in the target test set; the label represents the collection time period of the customer information; the collection time period comprises a historical time period and an updated time period; the updated time period is a time period from the last update time to the current time; the last update time represents the time when the content of the customer information was last updated; and the historical time period is a time period in which the end point coincides with the start point of the updated time period.

[0114] In this step, the customer data in the to-be-identified information is extracted, the customer data is divided into a target training set and a target test set, and the preset initial model is trained through the target training set to obtain a mature model, and the mature model is tested through the target test set to obtain a test accuracy rate, so as to determine whether the machine learning model can accurately determine the collection time period of the customer data based on the customer data.

[0115] In one preferred embodiment, training the preset initial model through the target training set to obtain a mature model comprises:

[0116] The first structured data in the target training set is used to train the random forest classification model in the initial model to obtain a mature forest classification model; wherein the first structured data refers to customer data with a structured data structure; and the structured data refers to data that can be represented and stored using a relational database and / or logically expressed using a two-dimensional table.

[0117] The text vector of the second structured data in the target training set is calculated, and the logic regression classification model in the initial model is trained through the text vector to obtain a mature regression classification model; wherein the second structured data refers to customer data with an unstructured data structure; the unstructured data refers to data without a fixed structure; and the text vector is used to represent the word frequency and inverse document frequency of the unstructured customer data.

[0118] The mature forest classification model and / or the mature regression classification model are summarized to obtain a mature model.

[0119] By way of example, the random forest classification model is an ensemble learning algorithm, which belongs to the Bagging type. The final result is obtained by voting or taking the average value, so that the result of the overall model has high accuracy and generalization performance. It can achieve good results, mainly due to the "random" and "forest", one of which makes it resistant to overfitting, and one of which makes it more accurate.

[0120] Therefore, by classifying the customer data of the structured data through the random forest classification model, the classification accuracy and generalization performance will be higher.

[0121] The logistic regression model is a machine learning method for solving binary classification (0 or 1) problems, which is used to estimate the possibility of something. For example, the possibility of a user purchasing a certain product, the possibility of a patient suffering from a certain disease, and the possibility of an advertisement being clicked by a user.

[0122] Therefore, by classifying the customer data of the unstructured data through the logistic regression model, the classification of the customer data based on the content of the customer data can be determined, and the accuracy of the classification of the unstructured data is ensured.

[0123] Term frequency and inverse document frequency refer to TF-IDF (Term Frequency-Inverse Document Frequency), wherein the term frequency (Term Frequency, TF) refers to the frequency of a given word appearing in the file. That is, the ratio of the number of times the word w appears in the document d count(w, d) and the total number of words in the document d size(d). The inverse document frequency (Inverse Document Frequency, IDF) is a measure of the general importance of a word. The IDF of a certain word can be obtained by dividing the total number of files by the number of files containing the word, and then taking the logarithm of the quotient. That is, the logarithm of the ratio of the total number of documents n and the number of files docs(w, D) in which the word w appears.

[0124] TF-IDF is a commonly used weighting technique for information retrieval and text mining. TF-IDF is a statistical method used to evaluate the importance of a word to a file set or a document in a corpus. The importance of a word increases in direct proportion to the number of times it appears in a file, but decreases in inverse proportion to its frequency in the corpus. Various forms of TF-IDF weighting are commonly used by search engines as a measure or rating of the relevance between files and user queries. The main idea of TF-IDF is that if a word or phrase appears frequently in an article and rarely in other articles, it is considered to have good class distinction ability and is suitable for classification. TF-IDF is actually: TF*IDF.

[0125] In a preferred embodiment, the mature model is tested by a target test set to obtain a test accuracy, comprising:

[0126] The first structure data in the target test set is labeled and predicted by a mature forest classification model in the mature model to obtain a forest verification label of the first structure data; wherein the forest verification label is a prediction result of the mature forest classification model in the label prediction of the first structure data;

[0127] The forest verification label of each first structure data is compared with the first label or the second label of the first structure data to determine the forest accuracy of the mature forest classification model:

[0128] The second structure data in the target test set is labeled and predicted by a mature regression classification model in the mature model to obtain a regression verification label of the second structure data; wherein the regression verification label is a prediction result of the mature regression classification model in the label prediction of the second structure data;

[0129] The regression verification label of each second structure data is compared with the first label or the second label of the second structure data to determine the regression accuracy of the regression classification model:

[0130] The test accuracy of the mature model is obtained according to the forest accuracy and the regression accuracy.

[0131] Further, the test accuracy of the mature model is obtained according to the forest accuracy and the regression accuracy, comprising:

[0132] The average value of the forest accuracy and the regression accuracy is calculated, and the average value is set as the test accuracy; or

[0133] The weighted value of the forest accuracy and the regression accuracy is calculated by a preset weight function, and the weighted value is set as the test accuracy; or

[0134] The forest accuracy and the regression accuracy are summarized to form an accuracy set, and the accuracy set is set as the test accuracy.

[0135] Optionally, if it is determined that the test accuracy is greater than the preset upper threshold, it is determined that the first data set and the second data set are portrait data, comprising:

[0136] If it is determined that the test accuracy includes a specified value, the specified value is compared with an upper specified threshold in the upper threshold; if it is determined that the specified value is higher than the upper specified threshold, it is determined that the test accuracy is greater than the upper threshold, and it is determined that the first data set and the second data set are portrait data; wherein the specified value is the average value or the weighted value of the forest accuracy and the regression accuracy; and / or

[0137] If it is determined that the test accuracy includes the forest accuracy, the forest accuracy is compared with an upper limit forest threshold in the upper limit threshold; if it is determined that the forest accuracy is higher than the upper limit forest threshold, the forest pass information is marked in the test accuracy; if it is determined that the test accuracy includes the regression accuracy, the regression accuracy is compared with an upper limit regression threshold in the upper limit threshold; if it is determined that the regression accuracy is higher than the upper limit regression threshold, the regression pass information is marked in the test accuracy; if it is determined that the test accuracy has the forest pass information and the regression pass information, it is determined that the test accuracy is greater than the upper limit threshold, and the first data set and the second data set are determined as the portrait data.

[0138] Optionally, if it is determined that the test accuracy is less than the preset lower limit threshold, the second data set is taken as the portrait data, and the failure information is generated, including:

[0139] If it is determined that the test accuracy includes the specified value, the specified value is compared with a lower limit specified threshold in the lower limit threshold; if it is determined that the specified value is lower than the lower limit specified threshold, it is determined that the test accuracy is less than the lower limit threshold, the second data set is taken as the portrait data, and the failure information is generated; wherein the specified value is the average or weighted value of the forest accuracy and the regression accuracy; and / or

[0140] If it is determined that the test accuracy includes the forest accuracy, the forest accuracy is compared with a lower limit forest threshold in the lower limit threshold; if it is determined that the forest accuracy is lower than the lower limit forest threshold, the forest pass information is marked in the test accuracy; if it is determined that the test accuracy includes the regression accuracy, the regression accuracy is compared with a lower limit regression threshold in the lower limit threshold; if it is determined that the regression accuracy is lower than the lower limit regression threshold, the regression pass information is marked in the test accuracy; if it is determined that the test accuracy has the forest pass information and the regression pass information, it is determined that the test accuracy is less than the lower limit threshold, the second data set is taken as the portrait data, and the failure information is generated.

[0141] S204: If it is determined that the test accuracy is greater than the preset upper limit threshold, the customer information is determined as the portrait data; wherein the portrait data is output to the intelligent recommendation system; the intelligent recommendation system is used to generate the user portrait of the customer corresponding to the customer information according to the portrait data.

[0142] In this step, if the test accuracy is greater than the upper limit threshold, it indicates that the machine learning model trained by the customer data can accurately identify which time period each customer data is in in the customer information, and then the content of the customer information in the historical time period and the content of the customer information in the updated time period are explained, even if they are different, it will not affect the portrait label of the user portrait, so the customer information can be directly taken as the portrait data for generating the user portrait.

[0143] Exemplarily, assuming that the upper limit threshold is 0.6 and the lower limit threshold is 0.3, if the test accuracy is greater than 0.6, it indicates that all the contents in the customer information are reliable, and can be output to the intelligent recommendation system, so that the intelligent recommendation system generates a user portrait according to the user data, to recommend products.

[0144] S205: If it is determined that the test accuracy is less than the preset lower limit threshold, the contents of the customer information in the update time period are determined as the portrait data, and invalidation information is generated; wherein the update time period is a time period from the last update time to the current time; the last update time represents the time when the contents of the customer information were last updated; the invalidation information represents the contents of the customer information in the historical time period are invalid; and the historical time period is a time period in which the end point coincides with the start point of the update time period.

[0145] Exemplarily, based on the above example, please refer to Figure 5 If the test accuracy is less than 0.3, it indicates that the contents of the customer information in the historical time period are unreliable and cannot be used for generating a user portrait, while the contents in the update time period are reliable and can be used for generating a user portrait, so that only the contents in the update time period can be sent to the intelligent recommendation system, and the intelligent recommendation system will regenerate a user portrait according to the contents in the update time period, to ensure the accuracy of the user portrait.

[0146] In this step, if the test accuracy is less than the lower limit threshold, it indicates that the machine learning model trained by the customer data cannot accurately identify which time period of the customer information each customer data is in, and thus it is indicated that the contents of the customer information in the historical time period have been invalidated, and therefore the contents of the customer information in the update time period need to be used as the portrait data for generating a user portrait.

[0147] Therefore, by identifying the customer data based on the machine learning model and then determining the contents of the customer information that can be used as the portrait data, the reliability for generating a user portrait is ensured, and thus the accuracy and timeliness of the obtained user portrait are ensured, the robustness of the intelligent recommendation system is enhanced, and the occurrence of customer dissatisfaction is avoided.

[0148] Embodiment 2:

[0149] Please refer to Figure 6 The present application provides a portrait data processing method, comprising:

[0150] S601: Collecting online transaction information and offline transaction information of a target customer;

[0151] The customer data corresponding to the structured label in the online transaction information and / or offline transaction information is used as the first structured data;

[0152] Customer data corresponding to unstructured tags in offline transaction information is used as the second structured data;

[0153] The customer data corresponding to the offline tags associated with unstructured tags in offline transaction information is used as the second structured data.

[0154] For example, the structured tags include one or more of the following: user login ID, transaction time, transaction type, transaction amount, transacting party account, transacting party account type, counterparty account type, transaction channel, and merchant ID.

[0155] Unstructured tags include: business address.

[0156] Offline labels include one or more of the following: country code, province code, and city code.

[0157] For example: Retrieve the user ID from the updated STATUS field, such as Figure 7 As shown, the user's income and expenditure details for the past two years are retrieved from the TransDetails table based on the user ID and the current time, and used as customer data.

[0158] like Figure 8 As shown, if the transaction channel is offline, then for offline transaction information, the necessary merchant information is collected through unstructured tags and offline tags to serve as the corresponding customer data.

[0159] S602: Identify the information to be identified in the database; wherein the database stores at least one customer information; the information to be identified is customer information that has not been updated for more than a preset update time; the customer information represents the attribute characteristics of the customer.

[0160] This step is the same as S201 in Example 1, so it will not be described again here.

[0161] S603: Extract customer data from the information to be identified, and divide the customer data into a target training set and a target test set; wherein, the target training set is used to train the machine learning model; the target test set is used to verify the trained machine learning model and predict the accuracy of the labels of the customer data in the target test set.

[0162] This step is the same as S202 in Example 1, so it will not be described again here.

[0163] S604: training the preset initial model through the target training set to obtain a mature model, and testing the mature model through the target test set to obtain a test accuracy rate; wherein the initial model comprises at least one machine learning model; the test accuracy rate represents the accuracy of the mature model in predicting the label of the customer data in the target test set; the label represents the collection time period of the customer information; the collection time period comprises a historical time period and an updated time period; the updated time period is a time period from the last update time to the current time; the last update time represents the time when the content of the customer information was last updated; and the historical time period is a time period in which the end point coincides with the start point of the updated time period.

[0164] This step is consistent with S203 in Embodiment 1, and thus is not described here.

[0165] S605: if it is determined that the test accuracy rate is greater than the preset upper threshold, the customer information is determined as the portrait data; wherein the portrait data is output to an intelligent recommendation system; and the intelligent recommendation system is used to generate a user portrait of the customer corresponding to the customer information according to the portrait data.

[0166] This step is consistent with S204 in Embodiment 1, and thus is not described here.

[0167] S606: if it is determined that the test accuracy rate is less than the preset lower threshold, the content of the customer information in the updated time period is determined as the portrait data, and invalid information is generated; wherein the updated time period is a time period from the last update time to the current time; the last update time represents the time when the content of the customer information was last updated; and the invalid information represents that the first data set has been invalidated.

[0168] This step is consistent with S205 in Embodiment 1, and thus is not described here.

[0169] S607: if it is determined that the test accuracy rate is less than or equal to the upper threshold and greater than or equal to the preset lower threshold, a weight value of the content of the customer information in the historical time period is reduced to obtain first correction information, and / or a weight value of the content of the customer information in the updated time period is increased to obtain second correction information.

[0170] The first correction information and the second correction information are summarized, or the content of the customer information in the historical time period and the second correction information are summarized, or the first correction information and the content of the customer information in the updated time period are summarized, to obtain the portrait data.

[0171] If the test accuracy is less than or equal to the upper limit threshold value and greater than or equal to the lower limit threshold value, it is indicated that the customer data in the first data set has reference value but is not important, and thus the first modified data set is obtained by reducing the weight value of the customer data in the first data set, for example, by inserting a low weight value into each customer data, and the low weight value is used to reduce the influence of the customer data in the first data set on the final user portrait when performing user portrait operation.

[0172] Exemplarily, if the test accuracy is less than or equal to the upper limit threshold value and greater than or equal to the lower limit threshold value, it is indicated that the customer data in the first data set has reference value but is not important, and thus the first modified data set is obtained by reducing the weight value of the customer data in the first data set, for example, by inserting a low weight value into each customer data, and the low weight value is used to reduce the influence of the customer data in the first data set on the final user portrait when performing user portrait operation.

[0173] In addition, the second modified data set can also be obtained by increasing the weight value of the customer data in the second data set, for example, by inserting a high weight value into each customer data, and the high weight value is used to increase the influence of the customer data in the second data set on the final user portrait when performing user portrait operation.

[0174] Optionally, if it is determined that the test accuracy is less than or equal to the upper limit threshold value and greater than or equal to the preset lower limit threshold value, the weight value of the first data set is reduced to obtain the first modified data set, and / or the weight value of the second data set is increased to obtain the second modified data set, including:

[0175] If it is determined that the test accuracy includes a specified value, the specified value is compared with the upper limit specified threshold value in the upper limit threshold value and the lower limit specified threshold value in the lower limit threshold value; if it is determined that the specified value is lower than the lower limit specified threshold value, it is determined that the test accuracy is less than the lower limit threshold value, the second data set is taken as the portrait data, and the invalidation information is generated; wherein the specified value is the average or weighted value of the forest accuracy and the regression accuracy; and / or

[0176] If it is determined that the test accuracy includes the forest accuracy, the forest accuracy is compared with a lower limit forest threshold value in the lower limit threshold value; if it is determined that the forest accuracy is lower than the lower limit forest threshold value, the forest pass information is marked in the test accuracy; if it is determined that the test accuracy includes the regression accuracy, the regression accuracy is compared with a lower limit regression threshold value in the lower limit threshold value; if it is determined that the regression accuracy is lower than the lower limit regression threshold value, the regression pass information is marked in the test accuracy; if it is determined that the test accuracy has the forest pass information and the regression pass information, it is determined that the test accuracy is less than the lower limit threshold value, the second data set is taken as the portrait data, and the failure information is generated.

[0177] In a preferred embodiment, the weight value of the content of the customer information in the historical time period is reduced to obtain first correction information, including:

[0178] The difference between the test accuracy and the lower limit threshold value is calculated to obtain a first difference value, and a first weight parameter corresponding to the first difference value is obtained through a preset first correction rule; wherein the first correction rule is used to represent the mapping relationship between the difference between different test accuracies and the lower limit threshold value and different weight parameters.

[0179] The first weight parameter is inserted into the content of the customer information in the historical time period to reduce the weight value of the content in the historical time period and obtain the first correction information.

[0180] For example, assuming that the test accuracy is 0.5 and the lower limit threshold value is 0.3, the first difference value is 0.2,

[0181] If the first weight parameter corresponding to the first difference value 0.2 is 0.6, then 0.6 is inserted into the content of the customer information in the historical time period to reduce the weight value of the content in the historical time period and obtain the first correction information.

[0182] So that the intelligent recommendation system reduces the weight of the content of the customer information in the historical time period from 1 to 0.6 when generating the user portrait, and further reduces the influence of the content in the historical time period on the user portrait.

[0183] In a preferred embodiment, the weight value of the content of the customer information in the historical time period is reduced to obtain first correction information, including:

[0184] The difference between the test accuracy and the lower limit threshold value is calculated to obtain a first difference value, and a first weight parameter corresponding to the first difference value is obtained through a preset first correction rule; wherein the first correction rule is used to represent the mapping relationship between the difference between different test accuracies and the lower limit threshold value and different weight parameters.

[0185] The second weight parameter is inserted into the content of the customer information in the update time period to increase the weight value of the content in the update time period and obtain second correction information.

[0186] Exemplarily, assuming that the test accuracy is 0.5 and the upper threshold is 0.6, the second difference is 0.1, and if the first weight parameter corresponding to the second difference 0.1 is 1.2, then 1.2 is inserted into the content of the customer information in the update time period to reduce the weight value of the content in the update time period and obtain second correction information.

[0187] So that the intelligent recommendation system changes the weight of the content of the customer information in the update time period from 1 to 1.2 when generating the user portrait, thereby improving the influence of the content in the historical time period on the user portrait.

[0188] Embodiment 3

[0189] Please refer to Figure 9 The application provides a portrait data generation device 9, comprising:

[0190] The acquisition module 92 is configured to identify to-be-identified information in a database, wherein the database stores at least one customer information, the to-be-identified information is customer information that has not been updated in content for more than a preset update time, and the customer information represents attribute characteristics of a customer.

[0191] The preprocessing module 93 is configured to extract customer data in the to-be-identified information and divide the customer data into a target training set and a target test set, wherein the customer data represents transaction behaviors and content in the past of the customer, the target training set is used for training a machine learning model, and the target test set is used for verifying the trained machine learning model and predicting the accuracy of labels of the customer data in the target test set.

[0192] The training and prediction module 94 is configured to train a preset initial model to obtain a mature model through the target training set and test the mature model through the target test set to obtain a test accuracy, wherein the initial model comprises at least one machine learning model, the test accuracy represents the accuracy of the mature model in predicting the labels of the customer data in the target test set, and the label represents a collection time period of the customer information, the collection time period comprises a historical time period and an update time period, the update time period is a time period from a last update time to a current time, the last update time represents a time when the content of the customer information was last updated, and the historical time period is a time period in which an end point coincides with a start point of the update time period.

[0193] The first comparison and identification module 95 is configured to determine that the customer information is the portrait data if it is determined that the test accuracy is greater than the preset upper threshold value; wherein the portrait data is output to the intelligent recommendation system; and the intelligent recommendation system is configured to generate a user portrait of the customer corresponding to the customer information according to the portrait data.

[0194] The second comparison and identification module 96 is configured to determine that the content of the customer information in the update time period is the portrait data and generate the invalid information if it is determined that the test accuracy is less than the preset lower threshold value; wherein the invalid information represents that the content of the customer information in the historical time period is invalid.

[0195] Optionally, the portrait data processing apparatus 9 further comprises:

[0196] The information collection module 91 is configured to collect online transaction information and offline transaction information of the target customer;

[0197] The customer data corresponding to the structured label in the online transaction information and / or the offline transaction information is taken as the first structured data;

[0198] The customer data corresponding to the unstructured label in the offline transaction information is taken as the second structured data;

[0199] The customer data corresponding to the offline label associated with the unstructured label in the offline transaction information is taken as the second structured data.

[0200] The third comparison and identification module 97 is configured to obtain the first correction information by reducing the weight value of the content of the customer information in the historical time period and / or obtain the second correction information by increasing the weight value of the content of the customer information in the update time period if it is determined that the test accuracy is less than or equal to the upper threshold value and greater than or equal to the preset lower threshold value.

[0201] The first correction information and the second correction information, or the content of the customer information in the historical time period and the second correction information, or the first correction information and the content of the customer information in the update time period are summarized to obtain the portrait data.

[0202] Embodiment 4:

[0203] Figure 10 is a block diagram of a computer device according to an exemplary embodiment. The computer device 10 can be a control circuit in the information management system 1, which is implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic elements, and is used to execute the portrait data generation method.

[0204] Specifically, the computer device 10 includes: a processor 102 and a memory 101 communicatively connected to the processor 102; the memory stores computer-executed instructions;

[0205] The processor executes computer execution instructions stored in memory 101 to implement the aforementioned portrait data generation method. The components of the portrait data generation device can be distributed across different computer devices. Computer device 10 can be a smartphone, tablet, laptop, desktop computer, rack server, blade server, tower server, or cabinet server (including standalone servers or server clusters composed of multiple application servers), etc. The computer device in this embodiment includes, but is not limited to, memory 101 and processor 102, which are interconnected via a system bus. Figure 3 As shown. It should be noted that, Figure 3 Only computer devices with components are shown; however, it should be understood that it is not required to implement all of the shown components, and more or fewer components may be implemented instead. In this embodiment, memory 101 (i.e., readable storage medium) includes flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, memory 101 may be an internal storage unit of the computer device, such as the hard disk or RAM of the computer device. In other embodiments, memory 101 may also be an external storage device of the computer device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. Of course, memory 101 may also include both internal storage units and external storage devices of the computer device. In this embodiment, the memory 101 is typically used to store the operating system and various application software installed on the computer device, such as the program code of the portrait data generation device in Embodiment 3. Furthermore, the memory 101 can also be used to temporarily store various types of data that have been output or will be output. The processor 102, in some embodiments, can be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 102 is typically used to control the overall operation of the computer device. In this embodiment, the processor 102 is used to run the program code stored in the memory 101 or process data, for example, to run the portrait data generation device to implement the portrait data generation method of the above embodiments.

[0206] The integrated modules implemented in the form of software function modules can be stored in a computer readable storage medium. The software function modules are stored in a storage medium and include a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute part of the steps of the method of various embodiments of the present application. It should be understood that the processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the application can be directly embodied as hardware processor execution, or executed by a combination of hardware and software modules in the processor. The memory can include a high-speed RAM memory, and can also include a non-volatile storage NVM, such as at least one disk memory, and can also be a U disk, a mobile hard disk, a read-only memory, a magnetic disk or an optical disk, etc.

[0207] To achieve the above-mentioned purpose, the present application also provides a computer readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card type memory (such as an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a disk, an optical disk, a server, an App application mall, etc., which stores computer execution instructions, and the program is executed by the processor 102 to realize the corresponding function. The computer readable storage medium of the embodiment is used to store computer execution instructions for realizing the image data generation method, and the processor 102 is executed to realize the image data generation method of the above-mentioned embodiment.

[0208] The storage medium can be realized by any type of volatile or non-volatile storage device or their combination, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0209] An example storage medium is coupled to the processor such that the processor can read information from, and can write information to, the storage medium. Of course, the storage medium can be a part of the processor. Consistent with the teachings provided herein, the processor can execute instructions embodied by computer code stored on the storage medium, and the storage medium can store computer code which, when executed by the processor, causes the processor to perform methods as described herein. The processor and the storage medium can be supplemented by, or incorporated in, ASICs (application- specific integrated circuits).

[0210] The present application provides a computer program product comprising a computer program which, when executed by a processor, implements the above-described image data generation method.

[0211] It should be noted that, in the present document, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element preceded by "comprises... a" does not, without more constraints, foreclose the existence of additional identical elements in the process, method, article, or apparatus that comprises the recited element.

[0212] Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the application be limited only by the scope of the claims, a properly construed, and any variations broadly included in the scope of the application. It is intended that the specification and examples be considered exemplary only, with the true scope and spirit of the application indicated only by the following claims.

[0213] It is to be understood that the application is not limited to the precise construction described and as shown in the attached figures, and that various modifications and changes can be made by those skilled in the art without departing from the scope of the application. The scope of the application is to be determined only by the appended claims.

Claims

1. An image data generating method characterized by comprising: The method comprises: identifying to-be-identified information in a database; wherein the database stores at least one customer information; the to-be-identified information is customer information that has not been updated for more than a preset update time; and the customer information represents the attribute characteristics of a customer; extracting customer data in the to-be-identified information, and dividing the customer data into a target training set and a target test set; wherein the customer data represents the transaction behavior and content of a customer in the past; the target training set is used for training a machine learning model; and the target test set is used for verifying the trained machine learning model and predicting the accuracy of the labels of the customer data in the target test set; training a preset initial model through the target training set to obtain a mature model, and testing the mature model through the target test set to obtain a test accuracy; wherein the initial model comprises at least one machine learning model; the test accuracy represents the accuracy of the mature model in predicting the labels of the customer data in the target test set; the labels represent the collection time period of the customer information; the collection time period comprises a historical time period and an update time period; the update time period is a time period from the last update time to the current time; the last update time represents the time when the content of the customer information was last updated; and the historical time period is a time period in which the end point coincides with the start point of the update time period; if it is determined that the test accuracy is greater than a preset upper threshold, determining that the customer information is portrait data; wherein the portrait data is output to an intelligent recommendation system; and the intelligent recommendation system is used for generating a user portrait of a customer corresponding to the customer information according to the portrait data; if it is determined that the test accuracy is less than a preset lower threshold, determining the content of the customer information in the update time period as the portrait data, and generating invalidation information; wherein the invalidation information represents the invalidation of the content of the customer information in the historical time period; training a preset initial model through the target training set to obtain a mature model, comprising: training a random forest classification model in the initial model through first structured data in the target training set to obtain a mature forest classification model; wherein the first structured data refers to customer data with a structured data structure; and the structured data refers to data that can be represented and stored using a relational database and / or logically expressed using a two-dimensional table; calculating a text vector of second structured data in the target training set, and training a logistic regression classification model in the initial model through the text vector to obtain a mature regression classification model; wherein the second structured data refers to customer data with an unstructured data structure; the unstructured data refers to data without a fixed structure; and the text vector is used to represent the word frequency and inverse document frequency of the unstructured customer data; summarizing the mature forest classification model and / or the mature regression classification model to obtain a mature model.

2. The image data generating method according to claim 1, characterized by, The method comprises: identifying to-be-identified information in a database; wherein the database stores at least one customer information; the to-be-identified information is customer information that has not been updated for more than a preset update time; and the customer information represents the attribute characteristics of a customer; acquiring, from the database, a last update time of target information of any customer information in the database as the target information; determining that the target information is pending information if it is determined that a time difference between the last update time and a current time exceeds the update time; determining that the pending information is to-be-identified information if it is determined that attribute information in the pending information has changed.

3. The image data generating method according to claim 2, characterized by, determining that the pending information is to-be-identified information if it is determined that attribute information in the pending information has changed, comprising: extracting attribute information of the target information in the collection time period; wherein the attribute information includes first attribute information and second attribute information; the first attribute information is used to represent attribute characteristics of a customer corresponding to the target information during a historical time period; and the second attribute information is used to represent attribute characteristics of the customer corresponding to the target information during an update time period; determining the target information as to-be-identified information if it is determined that the first attribute information is inconsistent with the second attribute information.

4. The image data generating method according to claim 1, characterized by, dividing the customer data into a target training set and a target test set, comprising: extracting at least one customer data in the to-be-identified information, and dividing at least one customer data into a first data set and a second data set; wherein the customer data represents transaction behaviors of a customer corresponding to customer information within a preset collection time period; the collection time period includes the historical time period and the update time period, and the end of the historical time period coincides with the start of the update time; the customer data in the first data set represents transaction behaviors of the customer within the historical time period; and the customer data in the second data set represents transaction behaviors of the customer within the update time period; labeling the customer data in the first data set with a first label and labeling the customer data in the second data set with a second label; wherein the first label represents customer information of the customer at the historical time period; and the second label represents customer information of the customer at the update time period; splitting the first data set into a first training set and a first test set, splitting the second data set into a second training set and a second test set, obtaining a target training set by aggregating the first training set and the second training set, and obtaining a target test set by aggregating the first test set and the second test set.

5. The image data generating method according to claim 1, characterized by, testing the mature model through the target test set to obtain a test accuracy, comprising: labeling the first structure data in the target test set through a mature forest classification model in the mature model to obtain a forest verification label of the first structure data; wherein the forest verification label is a prediction result of the mature forest classification model on the first structure data; comparing the forest verification label of each first structure data with the first label or the second label of the first structure data to determine a forest accuracy of the mature forest classification model: performing label prediction on the second structure data in the target test set by the mature regression classification model in the mature model to obtain a regression verification label of the second structure data; wherein the regression verification label is a prediction result of the mature regression classification model on the second structure data; comparing the regression verification label of each second structure data with the first label or the second label of the second structure data to determine a regression accuracy of the regression classification model; obtaining a test accuracy of the mature model according to the forest accuracy and the regression accuracy.

6. The image data generating method according to claim 5, characterized by, obtaining a test accuracy of the mature model according to the forest accuracy and the regression accuracy, comprising: calculating an average value of the forest accuracy and the regression accuracy, and setting the average value as the test accuracy; or calculating a weighted value of the forest accuracy and the regression accuracy by a preset weight function, and setting the weighted value as the test accuracy; or summing up the forest accuracy and the regression accuracy to form an accuracy set, and setting the accuracy set as the test accuracy.

7. The image data generating method according to claim 1, characterized by, Before identifying the to-be-identified information in the database, the method further comprises: collecting online transaction information and offline transaction information of a target customer; taking customer data corresponding to a structured label in the online transaction information and / or the offline transaction information as first structure data; taking customer data corresponding to an unstructured label in the offline transaction information as second structure data; taking customer data corresponding to an offline label associated with the unstructured label in the offline transaction information as the second structure data.

8. The image data generation method according to any one of claims 1 to 7, characterized by, After testing the mature model by the target test set to obtain a test accuracy, the method further comprises: if it is determined that the test accuracy is less than or equal to the upper threshold value and greater than or equal to a preset lower threshold value, reducing a weight value of content of the customer information in a historical time period to obtain first correction information, and / or increasing a weight value of content of the customer information in an update time period to obtain second correction information; summing up the first correction information and the second correction information, or summing up the content of the customer information in the historical time period and the second correction information, or summing up the first correction information and the content of the customer information in the update time period to obtain portrait data.

9. The image data generating method according to claim 8, characterized by, reducing the weight value of the content of the customer information in the historical time period to obtain the first correction information, comprising: calculating a first difference value between the test accuracy and the lower threshold value to obtain a first weight parameter corresponding to the first difference value by a preset first correction rule; wherein the first correction rule is used to represent a mapping relationship between different differences between the test accuracy and the lower threshold value and different weight parameters; inserting the first weight parameter into the content of the customer information in the historical time period to reduce the weight value of the content in the historical time period and obtain the first correction information.

10. The image data generating method according to claim 8, characterized by, increasing the weight value of the content of the customer information in the update time period to obtain the second correction information, comprising: A second difference value is calculated by subtracting the upper limit threshold value from the test accuracy rate, and a second weight parameter corresponding to the second difference value is obtained through a preset second correction rule; wherein the second correction rule is used to represent the mapping relationship between the difference value between different test accuracy rates and the upper limit threshold value and different weight parameters; The second weight parameter is inserted into the content of the customer information within the update time period to improve the weight value of the content within the update time period and obtain second correction information.

11. An image data generating apparatus characterized by comprising: Comprise: The acquisition module is used for identifying the to-be-identified information in the database; wherein the database stores at least one customer information; the to-be-identified information is customer information that has not been updated in content for more than a preset update time; and the customer information represents the attribute characteristics of the customer; The preprocessing module is used for extracting customer data in the to-be-identified information, and dividing the customer data into a target training set and a target test set; wherein the customer data represents the transaction behavior and content of the customer in the past; the target training set is used for training a machine learning model; and the target test set is used for verifying the trained machine learning model and predicting the accuracy of the label of the customer data in the target test set; The training and prediction module is used for training a preset initial model through the target training set to obtain a mature model, and testing the mature model through the target test set to obtain a test accuracy rate; wherein the initial model comprises at least one machine learning model; the test accuracy rate represents the accuracy of the mature model in predicting the label of the customer data in the target test set; the label represents the collection time period in which the customer information is located; the collection time period comprises a historical time period and an update time period; the update time period is a time period from the last update time to the current time; the last update time represents the time when the content of the customer information was last updated; and the historical time period is a time period in which the end point coincides with the start point of the update time period; The first comparison and identification module is used for determining that the customer information is portrait data if it is determined that the test accuracy rate is greater than a preset upper limit threshold value; wherein the portrait data is output to an intelligent recommendation system; and the intelligent recommendation system is used for generating a user portrait of the customer corresponding to the customer information according to the portrait data; The second comparison and identification module is used for determining that the content of the customer information within the update time period is portrait data and generating invalidation information if it is determined that the test accuracy rate is less than a preset lower limit threshold value; wherein the invalidation information represents the invalidation of the content of the customer information within the historical time period; The training and prediction module is specifically used for training a random forest classification model in the initial model through first structure data in the target training set to obtain a mature forest classification model; wherein the first structure data refers to customer data with a structured data structure; and the structured data refers to data that can be represented and stored using a relational database and / or logically expressed and implemented using a two-dimensional table. text vectors of the second structure data in the target training set are calculated, and a logistic regression classification model in an initial model is trained through the text vectors to obtain a mature regression classification model; wherein the second structure data refers to customer data with unstructured data; the unstructured data refers to data without fixed structure; the text vector is used to represent the word frequency and inverse document frequency of the customer data of the unstructured type; a mature model is obtained by summarizing the mature forest classification model and / or the mature regression classification model.

12. A computer device, comprising: comprise: a processor, and a memory connected to the processor in communication; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to realize the method in any one of claims 1 to 10.

13. A computer-readable storage medium, characterized in that, the computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by the processor to realize the method in any one of claims 1 to 10.

Images