Label automatic generation method and device, equipment and medium

Abstract

The embodiment of the invention relates to the technical field of data processing, and discloses an automatic tag generation method and device, equipment and a medium, and the method comprises the steps: collecting real-time data and historical data; obtaining a real-time processing result according to the real-time data and a label configuration platform; according to the historical data and a label configuration platform, obtaining a recalculation result; and obtaining a target label according to the real-time processing result and the re-calculation result. According to the method, business configuration and automatic label output are realized through mass data such as different user behavior logs, hundreds of financial institutions and various business system changes. Therefore, the efficiency is improved, the user experience is optimized, the operation cost is reduced, and data-driven intelligent development is realized.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of data processing, and in particular to a method, device, equipment and medium for automatically generating labels. Background Art

[0002] In the field of refined operation of Internet finance, a commonly used solution is that the operation personnel use historical experience and relevant data to deeply analyze and determine the labels they expect, and then the developers intervene, first recalculate and process the historical data according to the given logic, and further process the real-time data according to the same logic, and finally provide the label results obtained by these processes to the operation personnel for use. Its specific operation process covers the following key steps: First, the operation personnel fully refer to historical experience and the data they have, and analyze and sort out the labels that meet the operation needs and the corresponding processing logic; second, recalculate the historical data strictly according to the established logic; third, use the same logic to process the real-time data; fourth, merge and integrate the results obtained by processing the historical data with the results of the real-time data processing; fifth, the operation personnel use the final merged results to carry out refined operation operations such as pushing products to users. However, the existing technical solution has many significant shortcomings, as shown below: Because it involves more manual operations, it inevitably leads to a higher probability of human error, and in multiple links such as data processing and label generation, it is difficult to ensure consistency between various stages and parts, which poses a challenge to the accuracy and stability of operations. High maintenance cost: Because the technical solution is highly dependent on manual labor, frequent manual maintenance is required, which not only increases labor costs, but also reflects a strong dependence on technology, bringing a heavy burden to the operating company.

[0003] In summary, the existing technical solutions related to the refined operation of Internet finance urgently need to be improved and innovated to overcome the above-mentioned shortcomings and meet the industry's growing demand for efficient, flexible and precise operations. Summary of the invention

[0004] The purpose of the present invention is to at least provide a method, device, equipment and medium for automatic label generation, which can at least solve the problems of high error probability and high maintenance cost in the prior art, and at least achieve business configuration and automatic label output, thereby improving efficiency, optimizing user experience, reducing operating costs and achieving the beneficial effects of data-driven intelligent development.

[0005] In order to solve the above technical problems, at least one embodiment of the present invention provides a method for automatically generating labels, the method comprising:

[0006] Collect real-time and historical data;

[0007] According to the above real-time data and label configuration platform, real-time processing results are obtained;

[0008] According to the above historical data and label configuration platform, the above recalculation results are obtained;

[0009] According to the above real-time processing results and the above recalculation results, the target label is obtained.

[0010] At least one embodiment of the present invention further provides a device for automatically generating labels, the device comprising:

[0011] Collection module, used to collect real-time data and historical data;

[0012] A real-time result acquisition module is used to configure the platform according to the above real-time data and labels to obtain real-time processing results;

[0013] A recalculation result acquisition module is used to obtain the recalculation result according to the historical data and label configuration platform;

[0014] The label acquisition module is used to obtain the target label according to the above real-time processing result and the above recalculation result.

[0015] At least one embodiment of the present invention also provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the automatic label generation method.

[0016] At least one embodiment of the present invention further provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the above-mentioned automatic label generation method is implemented.

[0017] The automatic label generation method provided by the embodiment of the present invention realizes business configuration and automatic label output by analyzing massive data such as different user behavior logs, hundreds of financial institutions, and numerous business system changes, thereby improving efficiency, optimizing user experience, reducing operating costs, and realizing data-driven intelligent development. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] One or more embodiments are exemplarily described by the pictures in the corresponding drawings, and these exemplary descriptions do not constitute limitations on the embodiments.

[0019] Figure 1 is a flow chart of a method for automatically generating labels provided by an embodiment of the present invention;

[0020] Figure 2 is a flow chart of a method for automatically generating labels provided by another embodiment of the present invention;

[0021] Figure 3 is a schematic diagram of an automatic label generation device provided by another embodiment of the present invention;

[0022] Figure 4 It is a schematic structural diagram of an electronic device provided by another embodiment of the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. However, it can be understood by those skilled in the art that in the embodiments of the present invention, many technical details are proposed in order to enable the reader to better understand the present invention. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed in the present invention can also be implemented. The division of the following embodiments is for the convenience of description and should not constitute any limitation on the specific implementation of the present invention. The various embodiments can be combined and referenced with each other under the premise of no contradiction.

[0024] In order to facilitate understanding of the embodiments of the present invention, relevant contents about the automatic label generation method are first introduced here.

[0025] In the field of refined operation of Internet finance, a commonly used solution is that the operation personnel use historical experience and relevant data to deeply analyze and determine the labels they expect, and then the developers intervene, first recalculate and process the historical data according to the given logic, and further process the real-time data according to the same logic, and finally provide the label results obtained by these processes to the operation personnel for use. Its specific operation process covers the following key steps: First, the operation personnel fully refer to historical experience and the data they have, and analyze and sort out the labels that meet the operation needs and the corresponding processing logic; second, recalculate the historical data strictly according to the established logic; third, use the same logic to process the real-time data; fourth, merge and integrate the results obtained by processing the historical data with the results of the real-time data processing; fifth, the operation personnel use the final merged results to carry out refined operation operations such as pushing products to users. However, the existing technical solution has many significant shortcomings, as shown below: Because it involves more manual operations, it inevitably leads to a higher probability of human error, and in multiple links such as data processing and label generation, it is difficult to ensure consistency between various stages and parts, which poses a challenge to the accuracy and stability of operations. High maintenance cost: Because the technical solution is highly dependent on manual labor, frequent manual maintenance is required, which not only increases labor costs, but also reflects a strong dependence on technology, bringing a heavy burden to the operating company.

[0026] In summary, the existing technical solutions related to the refined operation of Internet finance urgently need to be improved and innovated to overcome the above-mentioned shortcomings and meet the industry's growing needs for efficient, flexible and precise operations.

[0027] In order to solve the technical problems of the high probability of errors and high maintenance costs in the prior art, the present invention proposes a method for automatically generating labels. The following is a detailed description of the implementation details of the method for automatically generating labels in this embodiment. The following content is only for the convenience of understanding the implementation details and is not necessary for implementing this solution.

[0028] The automatic tag generation method of this embodiment can be applied to electronic devices with communication, computing and data storage capabilities. The specific process can be as follows: Figure 1 As shown, including:

[0029] Step 101, collect real-time data and historical data.

[0030] Specifically, the above real-time data includes: business data, user behavior logs and business logs. The above historical data can be historical business data, historical user behavior logs, historical business logs, historical product usage feedback data, historical browsing record data, etc.

[0031] Among them, historical product usage feedback data includes:

[0032] User evaluation information: text evaluations and ratings given by users for Internet financial products they have used. By analyzing the sentiment of these evaluations and extracting key content, we can understand the aspects in which users are satisfied or dissatisfied. For example, some users commented that the approval process for a certain credit product is cumbersome.

[0033] Product usage duration and frequency data: For example, the duration and login frequency of a financial app used by users. These data can reflect the user’s dependence on the product and their stickiness.

[0034] Historical browsing data includes:

[0035] Information on browsed financial product pages: various financial product detail pages, information articles, and other content that users have browsed on Internet financial platforms.

[0036] Search keyword data: Keywords used by users when searching on the platform. These keywords directly reflect the users’ current financial needs or concerns.

[0037] Step 102, configure the platform according to the above real-time data and tags to obtain real-time processing results.

[0038] Specifically, step 102 further includes:

[0039] Step 102a, processing the above real-time data in real time to construct a real-time model.

[0040] In this embodiment, real-time processing may include the following:

[0041] Part I: Data Collection and Access:

[0042] In terms of business data: real-time acquisition of various business-related data occurring on the Internet financial platform, such as new financial product transaction order information (including product name, transaction amount, transaction time and other specific details), types of financial service business newly opened or closed by users, etc.

[0043] In terms of user behavior logs: collect the user's current operating behavior on the Internet financial platform, such as the financial product page that the user clicks to enter in real time, the length of time the user stays on the page, whether relevant operations are performed (such as filling in purchase information, submitting an application, etc.), and other behavioral details. Through means such as point-of-sale technology deployed on the client or web page, these behavior log data are transmitted in real time to the data processing center for subsequent analysis of the user's current interests and behavioral intentions.

[0044] In terms of business logs: they mainly cover some key logs generated by the operation of the business system itself, such as whether a certain financial service interface call is successful, the response time of the system processing business requests, and other information. Real-time access to these business logs helps monitor the real-time operation status of the business system, promptly discover potential problems or abnormal situations, and ensure the smooth development of the entire Internet financial business.

[0045] Part II: Data cleaning and preprocessing:

[0046] Unified format: Since the real-time data collected comes from various sources, the data formats are often not unified. For example, business data may be stored in a structured database table, while user behavior logs are more likely to be in semi-structured or unstructured text format. These data in different formats need to be standardized and unified into a format that is convenient for subsequent analysis, such as organizing various types of data into JSON format or data table format.

[0047] Denoising: In real-time data collection, there will inevitably be some interfering data or erroneous data. For example, users may have some abnormal behavior records due to misoperation, or individual fields may be lost or erroneous during data transmission in the business system. It is necessary to set reasonable data verification rules to identify and remove these "noise" data to ensure the quality of data for subsequent processing.

[0048] Missing value supplementation: Some fields in some real-time data may be missing. According to the specific data characteristics and business logic, some reasonable methods can be used to supplement them. For example, if the browsing time of a user is missing in the user behavior log, if the same user browses similar pages multiple times in a short period of time and has recorded time data, you can refer to these existing data to perform operations such as mean filling.

[0049] Part III: Real-time data analysis and feature extraction

[0050] Business data analysis: Analyze the current business development status based on the real-time access business data. For example, through real-time statistics on the transaction quantity and transaction amount of different financial products, understand the real-time popularity and market demand of each product, determine whether there are abnormal transaction fluctuations of any product, etc., extract business-related features such as the real-time transaction growth rate of products and the proportion of business volume in different regions, and prepare for the subsequent model construction.

[0051] User behavior analysis: Mining users' current behavior patterns and intentions from user behavior logs. For example, if a user browses multiple high-risk financial product pages in succession and then checks risk assessment-related content, we can extract user behavior characteristics such as "interested in high-risk products and concerned about risks." By analyzing and summarizing a large number of similar behaviors, we can more accurately grasp the user's current needs and psychological state.

[0052] Business system status analysis: Analyze the real-time performance and stability of the business system based on business logs. For example, collect statistics on the average response time and error call rate of each business interface, and extract features that reflect the health of the system, so as to timely discover potential bottlenecks or failure risks in the system and ensure the normal operation of the business.

[0053] Part 4, Data Integration and Association:

[0054] Cross-data source association: Associating and integrating data from different sources such as business data, user behavior logs, and business logs. For example, associating a user's real-time business transaction data with his behavior logs in the same time period can provide a more comprehensive understanding of the user's transaction background and motivations. For example, the user made a financial purchase transaction after comparing similar products many times. This cross-data source association analysis helps to build more accurate user portraits and business insights.

[0055] Feature fusion: The features extracted from different dimensions are integrated to form a comprehensive feature vector. For example, the real-time revenue features of the product in the business data, the user's attention to the product in the user behavior data, and the features reflecting the stability of the product's business system in the business log are integrated together and used as the input of the subsequent real-time model to improve the accuracy of the model's judgment of the real-time situation.

[0056] The above-mentioned real-time model can be a real-time user portrait model, a real-time risk warning model, a real-time product recommendation model, etc.

[0057] Step 102b, according to the above-mentioned label configuration platform, obtain the model conditions and label conditions corresponding to the real-time model.

[0058] Specifically, there are conditions configured by operators in the label configuration platform. According to these configured conditions, the model conditions and label conditions corresponding to the real-time model are obtained. Among them, the model conditions include: the conditions configured by operators, algorithm and parameter conditions, model evaluation and threshold conditions, etc. Further, the data range limitation conditions also include: time range conditions: operators can set time-related model conditions on the label configuration platform according to business needs, such as analyzing data for the past hour, day or specific business peak hours for real-time models. For example, for the real-time risk warning model, if the operator focuses on the risk situation during the daily trading peak period (such as 16:00-18:00 after the stock market closes), the time range can be set in this interval so that the model focuses on analyzing the business data, user behavior logs and business logs generated during this period, and accurately captures the abnormal risks that may occur during the frequent trading stage. Data source limitation conditions: clarify which business sectors or system modules the data based on the model comes from. For example, when an Internet financial platform includes a variety of financial product businesses (financial management, credit, insurance, etc.), operators can configure the model to only analyze real-time data related to financial management business, exclude interference from other business data, and allow the model to focus on specific business areas. For example, if the real-time product recommendation model only wants to optimize recommendations for credit products, then the data source can be limited to data sources related to the credit business.

[0059] Algorithm and parameter conditions also include selecting the applicable algorithm type: according to different real-time model goals and data characteristics, operators can configure specific algorithms as model conditions on the platform. For example, for a real-time user portrait model, if you want to build a portrait based on the similarity of user behavior, you may choose a clustering algorithm (such as K-Means clustering); for a real-time risk warning model, if you focus on making judgments based on existing rules, you can configure rule-based algorithm logic, or in order to improve prediction accuracy, choose machine learning classification algorithms such as logistic regression and random forests, and set corresponding algorithm parameters, such as the number of trees in a random forest, the maximum depth, etc. The different values ​​of these parameters will affect the performance and output of the model. And model training and update frequency parameters: determine how often the model is retrained to update its own parameters and prediction capabilities. For scenarios with extremely high real-time requirements, such as real-time risk warning models, you may need to set a shorter training update cycle, such as retraining every 10 minutes, so as to timely capture new risk characteristics brought about by changes in market environment, user behavior, etc.; for real-time product recommendation models, if the product update frequency is relatively low, the training update cycle can be appropriately extended to once a day, etc., and these parameters can be reasonably configured to balance the accuracy of the model and the consumption of computing resources.

[0060] Model evaluation and threshold conditions also include: Evaluation indicator selection: The operator will set the evaluation indicators used to measure the performance of the real-time model as model conditions. Common ones include accuracy (the proportion of correct judgments made by the classification model), recall (the proportion of actual positive examples predicted by the model), F1 value (an indicator that comprehensively considers accuracy and recall), etc. For example, for the real-time risk warning model, more attention is paid to the recall rate, hoping to detect all potential risk situations as much as possible, even if there will be some false positives, so the recall rate will be configured as a key evaluation indicator to regularly check whether the model has reached the expected performance level. And the warning or decision threshold setting: According to factors such as actual business needs and risk tolerance, determine the threshold conditions corresponding to the model output results. For example, in the real-time risk warning model, set the threshold for the amount of abnormal fund transfer. When the user transfers more than the set amount (such as 50,000 yuan) in a single time and meets other risk-related behavior characteristics at the same time, the model determines it as a high-risk situation and triggers an early warning; in the real-time product recommendation model, set the threshold for the recommendation score. Only financial products with a recommendation score higher than a certain score (such as 80 points) will be recommended to users, thereby controlling the quality and accuracy of the recommendation.

[0061] In addition, label conditions also include basic attribute label conditions, behavioral preference label conditions, and risk and demand label conditions.

[0062] Step 102c, determining the above model conditions and the above label conditions as real-time processing results.

[0063] Specifically, real-time processing results are obtained.

[0064] Step 103, obtaining the above-mentioned recalculation result according to the above-mentioned historical data and label configuration platform.

[0065] Specifically, step 103 also includes:

[0066] Step 103a, performing offline data processing on the above historical data to construct an offline model.

[0067] Among them, offline data processing includes:

[0068] Data collection and integration: including multi-source data aggregation and data format unification.

[0069] Multi-source data aggregation means collecting historical data related to Internet finance from different data sources, which may include business databases (storing various financial business transaction records, etc.), user information management systems (including user basic information, account information, etc.), historical log files (such as user operation behavior logs, system operation logs, etc.), etc. Bringing together these data scattered in various systems ensures data integrity and provides comprehensive materials for subsequent in-depth processing. For example, the purchase records in the financial business database, the identity information in the user registration system, and the log data of the user's operation behavior on the platform over the past period of time are all integrated.

[0070] Data format unification means that data formats from different sources often differ, some are in structured table form (such as data in a database), and some may be in semi-structured text format (such as log data). It is necessary to normalize the format of all collected data and convert them into a format that is convenient for subsequent analysis and processing, such as converting them into a structured data table or a specific parseable data format (such as JSON format, etc.), so that subsequent data processing operations can be carried out smoothly.

[0071] Data cleaning: including missing value processing, duplicate value removal and outlier processing, as follows:

[0072] Missing value processing: Check whether there are some missing fields in the historical data. Different processing methods can be used for missing values ​​according to the characteristics of the data and business logic. For example, if the user age field is partially missing, and the age data of other users in the same region and the same registration time interval are relatively concentrated, the mode of the age in the interval can be used to fill it; if the transaction amount field is missing, if the transaction has other related transaction information, the missing amount value can be estimated through related calculations, or the record can be directly marked as suspicious data, and then decide whether to remove it according to the situation.

[0073] Duplicate value removal: Find and remove duplicate data records in the data set to prevent these duplicate data from interfering with subsequent analysis and statistics. For example, in transaction record data, there may be multiple duplicate records of the same transaction due to system failures and other reasons. Duplicate values ​​are identified by comparing the values ​​of each field (such as transaction time, transaction amount, transaction parties and other key information are exactly the same), and only one record is retained to ensure the uniqueness of the data.

[0074] Outlier processing: Identify and process outliers in historical data. Outliers may be caused by data entry errors, system failures, or real extreme situations. For example, if the amount of a user's financial transaction is far beyond the transaction scale of normal users, further verification is required. If it is an error, the abnormal record should be corrected or removed; if it is a real special large transaction, its specialness can be marked separately and analyzed so that it can be treated specially in the subsequent model building process. Outliers can be found by setting reasonable data range thresholds (determined based on business experience and statistical analysis) and using statistical tools such as box plots.

[0075] Data transformation: including data standardization, data discretization and variable encoding, as follows:

[0076] Data standardization: In order to compare and analyze data of different magnitudes and ranges on the same scale, the relevant data needs to be standardized. For example, for user asset data (which may range from a few thousand yuan to millions of yuan) and transaction frequency data (which may range from a few times a month to dozens of times), common standardization methods such as Z-score standardization (converting data into a distribution with a mean of 0 and a standard deviation of 1) or Min-Max standardization (mapping data to the [0,1] interval) can be used to make these data comparable in subsequent data analysis, model building, and other links.

[0077] Data discretization: For some continuous data variables, if the subsequent analysis is more suitable for discrete categories, discretization operations are required. For example, the user's age data can be divided into different age categories according to a certain interval (such as 0-18 years old for teenagers, 19-30 years old for young people, 31-50 years old for middle-aged, etc.), or the user's account balance can be divided into different levels according to the amount (such as low balance, medium balance, high balance, etc.), which is convenient for statistical analysis and rule formulation based on category characteristics.

[0078] Variable encoding: For some data with categorical properties (such as user gender, financial product type, etc.), if it needs to be converted into a numerical form that can be better processed by computers, variable encoding is required. For example, using the One-Hot Encoding method, the user gender "male" is encoded as [1,0] and "female" is encoded as [0,1], so that these categorical variables can be correctly represented in the input of machine learning models, etc., to avoid model misinterpretation.

[0079] Data association and integration: including cross-data source association analysis and feature fusion and derivation, as follows:

[0080] Cross-data source association analysis: Connect and integrate data from different but related sources to mine more valuable information. For example, by associating the user's transaction record data with its corresponding behavior log data according to the user ID, we can know what operations the user performed before each transaction, which product pages the user browsed, etc., and have a more comprehensive understanding of the user's transaction motivation and behavior pattern. Effective connections between different data tables can be achieved through database association query operations (such as JOIN statements in SQL, etc.).

[0081] Feature fusion and derivation: Combine existing data features to generate new features through certain calculations, combinations, etc., to further enrich the data's expressiveness. For example, combining the user's transaction frequency and transaction amount can derive the feature of "average single transaction amount"; or combining the user's age, asset status, and risk preference information to construct a new feature such as "comprehensive risk tolerance index". These derived features can provide more valuable dimensions for subsequent data analysis, model building, etc.

[0082] Data storage and management: including the selection of storage methods and data version control and maintenance, as follows:

[0083] Choose a storage method: Choose a suitable storage method to save offline processed data based on factors such as the size of the data, frequency of use, and subsequent application scenarios. For massive amounts of historical data, a distributed storage system (such as Hadoop Distributed File System, HDFS) may be selected to ensure storage scalability and high data availability; if the amount of data is relatively small and the read and write speed requirements are high, a traditional relational database (such as MySQL, etc.) may also be a suitable choice. At the same time, for some intermediate result data or data that needs to be accessed quickly, caching technology (such as Redis, etc.) can also be used in combination to improve data access efficiency.

[0084] Data version control and maintenance: Establish a data version management mechanism to record data versions at different stages and after different processing, so as to facilitate the tracing of data change history and roll back to the previous available version when problems occur. At the same time, the stored data should be maintained regularly, such as checking the integrity of the data, cleaning up expired and useless data, etc., to ensure the effective management and continuous availability of data resources.

[0085] Step 103b, according to the above-mentioned label configuration platform, obtain the automatic recalculation conditions corresponding to the offline model.

[0086] Specifically, there are conditions configured by operators in the label configuration platform, and according to these configured conditions, the automatic recalculation conditions corresponding to the offline model are obtained.

[0087] Step 103c, obtaining the recalculation result according to the above-mentioned automatic recalculation conditions.

[0088] Specifically, the automated recalculation includes the following steps:

[0089] First, the data preparation stage:

[0090] Determine the scope of recalculation: First, clarify the scope of historical data that needs to be recalculated, which may be defined based on the labels set by the operator and the corresponding processing logic. For example, if the operator determines that the user transaction data and behavior data related to certain specific financial products in the past year need to be recalculated to re-evaluate the accuracy of the relevant labels, the system will select the corresponding data related to these products within the time period as the basic data set for recalculation, ensuring that the recalculation work focuses on the target data range and avoids unnecessary waste of resources.

[0091] Data extraction and loading: extract historical data of a certain scope from the corresponding storage system (such as database, data warehouse, etc.) and load it into the recalculation working environment. This process involves interacting with different storage media, reading the data through appropriate data access interfaces (such as database query statements, etc.), and then loading it into memory or temporary computing space in an appropriate format (such as structured data table form) to provide a data basis for subsequent recalculation operations and ensure that the data can be smoothly accessed and processed by the recalculation program.

[0092] Secondly, during the recalculation rule configuration phase:

[0093] Analyze operational needs: The automated recalculation system needs to accurately understand the labels and corresponding processing logic set by the operators and convert them into computer-executable recalculation rules. For example, the operators proposed to determine the label of "active financial management users" based on the frequency, amount and redemption of users' purchases of financial products. The recalculation system needs to analyze how to calculate and judge users who meet the label definition from relevant data fields (such as the number of purchases, the amount of each purchase, the redemption time, etc.), and clearly convert these operational needs into specific recalculation rules and conditional judgment statements.

[0094] Rule parameter setting: Accurately set some parameters involved in the recalculation rules. For example, when calculating whether the frequency of users purchasing wealth management products meets the "active" standard, it is necessary to set a specific frequency threshold (such as more than or equal to 5 purchases in the past year is considered active), or when judging whether the redemption situation is normal, set parameters such as the reasonable range of redemption time within the product period. The setting of these parameter values ​​directly affects the accuracy of the recalculation results, and is usually determined comprehensively based on business experience, historical data analysis, and operational goals.

[0095] Then, the automation calculation phase:

[0096] Data cleaning and preprocessing (reconfirmation): Before the formal recalculation, although the previous offline data processing may have been done, it is still necessary to clean and preprocess the loaded data again to ensure that the data quality meets the recalculation requirements. For example, check whether there are new missing values ​​or duplicate values, further standardize the data format, etc., eliminate data problems that may interfere with the accuracy of the recalculation, and lay a good foundation for accurate calculation.

[0097] Calculation according to rules: According to the configured recalculation rules and set parameters, the prepared data is automatically calculated. This may involve various data operations, logical judgments, and statistical analysis. For example, according to the recalculation rules of the "active financial management users" label mentioned above, the system will count the actual number of times each user purchases financial products in the corresponding time period, compare it with the set frequency threshold, and analyze whether the redemption situation is within a reasonable range. Through a series of calculations and judgments, it is determined which users meet the definition of the label. The entire calculation process is automatically executed by the recalculation program without manual intervention, which greatly improves efficiency and accuracy.

[0098] Finally, the result verification and feedback stage:

[0099] Verification of result accuracy: Compare and verify the recalculated results with the previous historical results or expected results to check whether there are any differences and whether the differences are within a reasonable range. The quality of the recalculated results can be measured by some statistical indicators (such as accuracy, recall, etc., if applicable), or by manually sampling some typical cases to intuitively judge the rationality of the results. For example, for the list of "active financial management users" obtained by recalculation, select some known active users for verification to see if they are correctly identified. If a large deviation is found in the results, it is necessary to check whether it is caused by problems with the recalculation rules, data quality problems, or other reasons.

[0100] Feedback and adjustment: If problems are found in the recalculation results during the verification process, timely feedback will be given to the corresponding links (such as recalculation rule configuration, data preparation, etc.) for targeted adjustments and optimization. For example, if a parameter in the recalculation rule is set unreasonably and the result is inaccurate, the parameter will be reset and adjusted and recalculated again until the recalculation result meets expectations, ensuring the accuracy and reliability of the recalculation data and labels finally provided to the operators.

[0101] Step 104, obtaining a target label according to the above real-time processing result and the above recalculation result.

[0102] Specifically, the real-time processing result and the recalculation result are combined to obtain the target label. The above-mentioned combination method includes: any one of simple splicing, dictionary combination, weight distribution combination and conditional judgment combination.

[0103] In some embodiments, after obtaining the recalculation result according to the automatic recalculation condition, the method further includes:

[0104] First, the above recalculation results are cross-validated; secondly, the above recalculation results are updated according to the cross-validation results.

[0105] This embodiment improves the accuracy of the recalculation result by verifying the recalculation result.

[0106] The following is a specific example to illustrate the automatic label generation method provided by the present invention: Figure 2 As shown in the figure, first, a large amount of business data, user behavior logs, and business logs are collected. Secondly, real-time models of different dimensions are established through real-time processing. At the same time, historical data is processed according to the same logic to produce corresponding offline models. After that, according to the conditions configured by the operator on the label configuration platform, the corresponding model conditions and label conditions are generated, and the historical data is automatically recalculated. Finally, the recalculation results and the real-time processing results are merged into the required labels.

[0107] This embodiment realizes business configuration and automatic output labeling by analyzing massive data such as different user behavior logs, hundreds of financial institutions, and numerous business system changes, thereby improving efficiency, optimizing user experience, reducing operating costs, and realizing data-driven intelligent development.

[0108] Another embodiment of the present invention relates to an automatic label generation device. The implementation details of the automatic label generation of this embodiment are specifically described. The following content is only for the convenience of understanding the implementation details, which is not necessary for the implementation of this solution. The schematic diagram of the automatic label generation device of this embodiment can be as follows: Figure 3 As shown, it includes a collection module 301, a real-time result acquisition module 302, a complex calculation result acquisition module 303 and a label acquisition module 304. Specifically,

[0109] The collection module 301 is used to collect real-time data and historical data.

[0110] The real-time result acquisition module 302 is used to obtain real-time processing results according to the above real-time data and label configuration platform.

[0111] The recalculation result acquisition module 303 is used to obtain the recalculation result according to the historical data and the tag configuration platform.

[0112] The label acquisition module 304 is used to obtain a target label according to the above real-time processing result and the above recalculation result.

[0113] It is worth mentioning that all modules involved in this embodiment are logic modules. In practical applications, a logic unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. In addition, in order to highlight the innovative part of the present invention, this embodiment does not introduce units that are not closely related to solving the technical problem proposed by the present invention, but this does not mean that there are no other units in this embodiment.

[0114] Another embodiment of the present invention relates to an electronic device, such as Figure 4 As shown, it includes: at least one processor 401; and a memory 402 that is communicatively connected to the at least one processor 401; wherein the memory 402 stores instructions that can be executed by the at least one processor 401, and the instructions are executed by the at least one processor 401, so that the at least one processor 401 can execute the automatic label generation method in the above-mentioned embodiments.

[0115] Among them, the memory and the processor are connected in a bus manner, and the bus may include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors and memories together. The bus can also connect various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be one element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. The data processed by the processor is transmitted on a wireless medium via an antenna, and further, the antenna also receives data and transmits the data to the processor.

[0116] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory can be used to store data used by the processor when performing operations.

[0117] Another embodiment of the present invention relates to a computer-readable storage medium storing a computer program, which implements the above method embodiment when executed by a processor.

[0118] That is, those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including a number of instructions for a device (which can be a single-chip microcomputer, chip, etc.) or a processor (processor) to perform all or part of the steps of the above-mentioned method of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as: ROM), random access memory (Random Access Memory, referred to as: RAM), disk or optical disk and other media that can store program codes.

[0119] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present invention, and that in actual applications, various changes may be made in form and detail without departing from the spirit and scope of the present invention.

Claims

1. A method for automatically generating labels, characterized in that: The method comprises: Collect real-time and historical data; According to the real-time data and label configuration platform, obtain real-time processing results; According to the historical data and label configuration platform, a recalculation result is obtained; A target label is obtained according to the real-time processing result and the recalculation result.

2. The method for automatically generating labels according to claim 1, characterized in that: The real-time data includes: business data, user behavior logs and business logs.

3. The method for automatically generating labels according to claim 1, characterized in that: The real-time processing result is obtained according to the real-time data and label configuration platform, including: Processing the real-time data in real time to build a real-time model; According to the label configuration platform, a model condition and a label condition corresponding to the real-time model are obtained; The model condition and the label condition are determined as real-time processing results.

4. The method for automatically generating labels according to claim 1, characterized in that: The obtaining of the recalculation result according to the historical data and the label configuration platform includes: Performing offline data processing on the historical data to construct an offline model; According to the label configuration platform, an automatic recalculation condition corresponding to the offline model is obtained; According to the automatic recalculation condition, a recalculation result is obtained.

5. The method for automatically generating labels according to claim 1, characterized in that: The step of obtaining a target label according to the real-time processing result and the recalculation result includes: The real-time processing result and the recalculation result are combined to obtain a target label.

6. The method for automatically generating labels according to claim 5, characterized in that: The merging method includes: any one of simple splicing, dictionary merging, weight distribution merging and conditional judgment merging.

7. The method for automatically generating labels according to claim 4, characterized in that: After obtaining the recalculation result according to the automatic recalculation condition, the method further includes: Cross-validating the recalculation results; The recalculation result is updated according to the cross-validation result.

8. A label automatic generation device, characterized in that: The device comprises: Collection module, used to collect real-time data and historical data; A real-time result acquisition module is used to obtain real-time processing results according to the real-time data and label configuration platform; A recalculation result acquisition module, used to obtain the recalculation result according to the historical data and the tag configuration platform; The label acquisition module is used to obtain the target label according to the real-time processing result and the recalculation result.

9. An electronic device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the automatic label generation method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the automatic label generation method according to any one of claims 1 to 7 is implemented.

Images