User big data portrait generation method based on transaction scene

Abstract

The invention relates to the technical field of computers, and discloses a user big data portrait generation method based on a transaction scene, and the method comprises the steps: obtaining transaction data of a user from a plurality of different data sources, carrying out the preprocessing of the obtained transaction data, obtaining transaction scene feature data based on the data sources, and generating a user big data portrait. Modeling by using a pre-trained modeling model in combination with the preprocessed transaction data and the transaction scene characteristic data to generate a preliminary user portrait; applying the generated user portrait to a specific business scene, obtaining a social network relationship of the user in different specific business scenes, optimizing the preliminary user portrait based on the social network relationship, and obtaining a target user portrait; and calling a pre-established dynamic monitoring mechanism for user behaviors, periodically acquiring comment feedback data corresponding to the user in the specific business scene based on the dynamic monitoring mechanism, and periodically updating the corresponding target user portrait based on the comment feedback data. According to the invention, a finer user portrait can be constructed.

Description

Technical Field

[0001] The present application relates to the technical field of computers, and in particular to a method for generating user big data portraits based on transaction scenarios. Background Art

[0002] With the rapid development of Internet technology, online transactions such as e-commerce and mobile payments are becoming more frequent, generating massive amounts of transaction data. These data not only include users’ consumption records, but also contain important information such as their consumption habits, preferences, and credit status. Traditional data analysis methods can only provide a rough user portrait and cannot meet the needs of precision marketing and personalized services in the modern business environment.

[0003] In recent years, the development of big data and artificial intelligence technology has made it possible to solve this problem. By using advanced data mining and machine learning algorithms, valuable information can be extracted from complex transaction data to build a more sophisticated and multi-dimensional user portrait. However, the existing user portrait generation technology still has the following shortcomings: the data source is single, which makes it difficult to fully reflect the real situation of users; there is a lack of effective tracking mechanism for the dynamic changes of user behavior, resulting in poor timeliness of user portraits; and the differences in user behavior patterns in different transaction scenarios are ignored, which affects the accuracy of the portrait.

[0004] From the above, we can see that the problem of how to build a more detailed user portrait still needs to be solved. Summary of the invention

[0005] In order to construct a more detailed user portrait, the present application provides a method for generating user big data portraits based on transaction scenarios.

[0006] In the first aspect, the present application provides a method for generating a user big data portrait based on a transaction scenario, which adopts the following technical solution: A method for generating a user big data portrait based on a transaction scenario comprises: obtaining transaction data of a user from a plurality of different data sources, preprocessing the acquired transaction data, obtaining corresponding transaction scenario characteristic data based on the data source, combining the preprocessed transaction data with the corresponding transaction scenario characteristic data to perform modeling using a pre-trained modeling model, and generating a corresponding preliminary user portrait, wherein the data source comprises an e-commerce platform, a mobile payment platform, and a bank transaction record; applying the generated user portrait to a specific business scenario, obtaining the social network relationship of the user in different specific business scenarios, optimizing the preliminary user portrait based on the social network relationship, and obtaining a corresponding target user portrait, wherein the specific business scenario comprises personalized recommendation, risk assessment, and market analysis; calling a pre-established dynamic monitoring mechanism for user behavior, regularly obtaining comment feedback data corresponding to the user in a specific business scenario based on the dynamic monitoring mechanism, and regularly updating the corresponding target user portrait based on the comment feedback data.

[0007] By adopting the above technical solutions, the initial user portrait generated by combining multi-source data and transaction scenario characteristic data is more comprehensive and accurate; the dynamic monitoring mechanism ensures that the user portrait can timely reflect the latest changes in user behavior, improving the real-time and effectiveness of the portrait; optimizing the user portrait based on social network relationships and comment feedback data can provide more personalized services; in specific business scenarios such as personalized recommendations, risk assessment, and market analysis, more detailed user portraits can provide more accurate decision support, improve business results and user satisfaction. Through these solutions, more detailed and dynamic user portraits can be constructed, which can not only improve the accuracy and real-time performance of user portraits, but also better meet the needs of different business scenarios.

[0008] Optionally, in the process of optimizing the preliminary user portrait based on the social network relationship, the method also includes: obtaining the user's evaluation data on the recommended products in different specific business scenarios and opinion data on the market analysis results based on the comment feedback data; extracting the user's sentiment tendency data in the evaluation data and the opinion data through a natural language processor; and optimizing the preliminary user portrait based on the sentiment tendency data and the social network relationship.

[0009] By adopting the above technical solutions, the user's actual feelings and social behaviors can be more accurately reflected through sentiment tendency data and social network relationships, making the user portrait more accurate and comprehensive; by regularly obtaining and analyzing user comment feedback data, the user portrait can be dynamically adjusted to ensure that it always reflects the user's latest behavior and preferences, thereby improving the real-time and effectiveness of the service.

[0010] Optionally, the method also includes: using time series analysis technology to obtain the user's corresponding time change trend from the user's transaction data, wherein the time change trend includes transaction frequency, transaction amount, transaction time, and login frequency; based on the time change trend, determining whether there is periodic behavior or long-term continuous trend; if so, setting a corresponding threshold trigger, and generating warning data if the transaction data exceeds the threshold trigger.

[0011] By adopting the above technical solutions and using time series analysis technology, it is possible to capture the temporal trend of user behavior, identify periodic behavior and long-term continuous trends, and improve the ability to predict user behavior; by setting threshold triggers, when user transaction data exceeds the normal range, early warning data can be generated in a timely manner to help enterprises and financial institutions quickly discover potential risks. Through early warning data, risk assessment strategies can be adjusted in a timely manner, risk management processes can be optimized, and potential credit risks and fraud risks can be reduced.

[0012] Optionally, the method also includes: retrieving a pre-trained quality scoring model and pre-processed transaction data; inputting the pre-processed transaction data into the quality scoring model for quality scoring to obtain corresponding scoring data; comparing the scoring data with a pre-set scoring threshold, and if the scoring data is less than the scoring threshold, re-obtaining the corresponding transaction data.

[0013] By adopting the above technical solution, the pre-processed transaction data is scored through the quality scoring model to ensure the integrity and accuracy of the data and improve the reliability of the data; high-quality data can generate more accurate user portraits, improve the reliability and practicality of user portraits, and thus provide more accurate decision support in business scenarios such as personalized recommendations, risk assessment, and market analysis.

[0014] Optionally, the method also includes: obtaining the user's real-time behavior data in real time, and obtaining context information corresponding to the real-time behavior data, wherein the real-time behavior data includes but is not limited to browsing records, search records, click records, and purchase records, and the context information includes but is not limited to time, location, device type, and network environment; adjusting corresponding recommended content based on the real-time behavior data and the context information; obtaining corresponding recommendation feedback data based on the recommended content, and adjusting the recommendation strategy in combination with the target user portrait and the recommendation feedback data.

[0015] By adopting the above technical solutions, by obtaining real-time user behavior data and context information in real time, it is possible to dynamically adjust the recommended content, ensure the accuracy and timeliness of the recommendation, and improve the user experience; by obtaining recommendation feedback data and combining it with the target user portrait, the recommendation strategy can be continuously optimized to improve the performance of the recommendation system and user satisfaction.

[0016] Optionally, the method also includes: obtaining previous interaction record data corresponding to the user, wherein the previous interaction record data includes customer service conversation records, complaint records, and service request records; identifying preference data and demand data corresponding to the user based on the previous interaction record data; adjusting the recommendation strategy based on the preference data and demand data, and recommending the latest recommended content to the user based on the adjusted recommendation strategy.

[0017] By adopting the above technical solution and analyzing the user's previous interaction records, it is possible to more accurately identify the user's preferences and needs and provide more personalized services; based on the user's preference and demand data, adjust the recommendation strategy to ensure that the recommended content is more in line with the user's actual needs, and by providing recommended content that better meets user needs, enhance user satisfaction and loyalty and improve the overall user experience.

[0018] Optionally, the method also includes: retrieving key features corresponding to user-related data, the key features including transaction frequency, transaction amount, transaction time, login frequency, number of interactions, number of complaints, and customer service conversation records; determining the user's corresponding churn level based on the key features in combination with the time series; and generating corresponding retention measures based on the churn level, wherein the retention measures include personalized discounts, customized services, proactive contact, and satisfaction surveys.

[0019] By adopting the above technical solution, by retrieving the key features of users and combining them with time series analysis, it is possible to accurately predict the user churn risk and determine the user churn level; based on the user churn level, generate personalized retention measures, such as personalized discounts, customized services, proactive contact and satisfaction surveys, to improve the retention effect, and through effective retention measures, reduce user churn and improve customer relationships.

[0020] In the second aspect, the present application provides a user big data portrait generation device based on transaction scenarios, which adopts the following technical solutions: A user big data portrait generation device based on transaction scenarios, comprising: A transaction scenario characteristic data acquisition module acquires user transaction data from multiple different data sources, and pre-processes the acquired transaction data, and acquires corresponding transaction scenario characteristic data based on the data source, and combines the pre-processed transaction data with the corresponding transaction scenario characteristic data to perform modeling using a pre-trained modeling model to generate a corresponding preliminary user portrait, wherein the data source includes an e-commerce platform, a mobile payment platform, and bank transaction records; A target user portrait acquisition module applies the generated user portrait to specific business scenarios, obtains the social network relationships of users in different specific business scenarios, and optimizes the preliminary user portrait based on the social network relationships to obtain the corresponding target user portrait, wherein the specific business scenarios include personalized recommendations, risk assessment, and market analysis; The target user portrait update module calls a pre-established dynamic monitoring mechanism for user behavior, and regularly obtains user comment feedback data corresponding to specific business scenarios based on the dynamic monitoring mechanism, and regularly updates the corresponding target user portrait based on the comment feedback data.

[0021] In the third aspect, the present application provides a method for generating a user big data portrait based on a transaction scenario, which adopts the following technical solution: A user big data portrait generation system based on transaction scenarios includes a processor in which a program of any one of the above-mentioned user big data portrait generation methods based on transaction scenarios is running.

[0022] In a fourth aspect, the present application provides a storage medium, which adopts the following technical solution: A storage medium storing a program for any one of the above-mentioned methods for generating a user big data portrait based on a transaction scenario.

[0023] In summary, the present application includes at least one of the following beneficial technical effects: By combining multi-source data and transaction scenario characteristic data, the generated preliminary user portrait is more comprehensive and accurate; the dynamic monitoring mechanism ensures that the user portrait can promptly reflect the latest changes in user behavior, improving the real-time and effectiveness of the portrait; optimizing the user portrait based on social network relationships and comment feedback data can provide more personalized services and enhance user experience.

[0024] In specific business scenarios such as personalized recommendations, risk assessment, and market analysis, more detailed user portraits can provide more accurate decision-making support, improve business results and user satisfaction; through emotional tendency data and social network relationships, they can more accurately reflect users' actual feelings and social behaviors, making user portraits more accurate and comprehensive.

[0025] Through time series analysis technology, we can capture the temporal trend of user behavior, identify periodic behavior and long-term trends, and generate early warning data in a timely manner to help enterprises and financial institutions quickly discover potential risks and optimize risk management; we can score the pre-processed transaction data through a quality scoring model to ensure the integrity and accuracy of the data and improve the reliability of the data. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1It is a flowchart of a method for generating a user big data portrait based on a transaction scenario according to an exemplary embodiment.

[0027] Figure 2 It is a structural block diagram of a method and device for generating a user big data portrait based on a transaction scenario according to an exemplary embodiment. DETAILED DESCRIPTION

[0028] Embodiments of the present application are described in detail below, examples of which are shown in the accompanying drawings.

[0029] In the description of this specification, the description with reference to the terms "certain embodiments", "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0030] The present application embodiment discloses a method for generating a user big data portrait based on a transaction scenario, referring to Figure 1 ,include: S100, obtain the user's transaction data from multiple different data sources, and pre-process the acquired transaction data, obtain corresponding transaction scenario characteristic data based on the data source, combine the pre-processed transaction data with the corresponding transaction scenario characteristic data using a pre-trained modeling model to perform modeling, and generate a corresponding preliminary user portrait.

[0031] Among them, the user's transaction data is obtained from multiple different data sources (such as e-commerce platforms, mobile payment platforms, bank transaction records), and the data sources may include the user's purchase records, payment records, bank account transaction records, etc.

[0032] The process of preprocessing transaction data specifically includes: data cleaning: removing invalid or erroneous data and processing missing values; format conversion: converting data into a unified format for subsequent processing; outlier detection: identifying and processing outliers to ensure data accuracy.

[0033] It should be pointed out here that the corresponding transaction scenario characteristic data, such as transaction time, transaction amount, transaction frequency, purchase category, etc., are extracted from the data source; by using the pre-trained modeling model and combining the pre-processed transaction data and transaction scenario characteristic data for modeling, a preliminary user portrait can be generated. The preliminary user portrait includes the user's consumption habits, preferences, behavior patterns, etc.

[0034] In the scheme of the embodiment of the present application, the training process of the modeling model is simply summarized as follows: obtaining the user's transaction data from multiple data sources (such as e-commerce platforms, mobile payment platforms, and bank transaction records), and preprocessing these data, including data cleaning, format conversion, and outlier detection. Then, extract key features in the transaction data, such as transaction frequency, transaction amount, transaction time, login frequency, etc. Next, use the preprocessed transaction data and the extracted feature data as input, and use machine learning algorithms (such as logistic regression, random forest, neural network, etc.) for training to generate a modeling model that can accurately reflect user behavior patterns. Optimize model performance through cross-validation and parameter adjustment to ensure that the model has high accuracy and generalization ability when generating preliminary user portraits. Finally, apply the trained modeling model to the scheme of the present application.

[0035] Based on the above steps, first, obtaining data from multiple data sources can ensure the comprehensiveness and accuracy of user portraits; second, ensuring the integrity and accuracy of data through preprocessing can improve the reliability of modeling; finally, by combining the characteristic data of transaction scenarios, user portraits can be more in line with actual business scenarios.

[0036] S110, applying the generated user portrait to a specific business scenario, obtaining the social network relationships of the user in different specific business scenarios, optimizing the preliminary user portrait based on the social network relationships, and obtaining the corresponding target user portrait.

[0037] Among them, it is necessary to first determine the specific business scenarios, such as personalized recommendations, risk assessment, market analysis, etc., and apply the generated preliminary user portraits to the selected business scenarios. For example, in the personalized recommendation scenario, user portraits are used to recommend products; in the risk assessment scenario, user portraits are used to assess the user's credit risk; in the market analysis scenario, user portraits are used to analyze market trends.

[0038] Secondly, obtain the user's social network relationships from social platforms, forums, communities and other channels. Social network relationships include friend lists, follow lists, interaction records, etc. The obtained social network relationships are integrated with user portrait data to form more comprehensive user information.

[0039] Use graph analysis technology or social network analysis algorithms to analyze users' social network relationships and identify users' important social nodes and social behavior patterns. Then, extract key features from social network relationships, such as the number of friends, interaction frequency, social influence, etc., combine the extracted social features with the preliminary user portrait, optimize the preliminary user portrait, and generate the target user portrait. For example, if a user has a high influence in the social network, add an "opinion leader" label to the user portrait and adjust the recommendation strategy to better meet his needs.

[0040] Through the above steps, not only can the accuracy and comprehensiveness of user portraits be improved, but more precise services can also be provided in multiple business scenarios, improving user experience and business results.

[0041] S120, calling a pre-established dynamic monitoring mechanism for user behavior, regularly obtaining user comment feedback data corresponding to specific business scenarios based on the dynamic monitoring mechanism, and regularly updating the corresponding target user portrait based on the comment feedback data.

[0042] The streaming computing framework can be used to collect user transaction data and behavior data in real time, and then design and deploy data collection points to ensure that the user's behavior data in different business scenarios can be captured in real time, such as browsing history, search history, click history, purchase history, etc. In addition, monitoring rules need to be set to define which behavior data needs to be collected, as well as the frequency and time interval of collection; through the dynamic monitoring mechanism, user behavior can be monitored in real time to ensure the timeliness and accuracy of the data.

[0043] Through a dynamic monitoring mechanism, we obtain users' comment feedback data in specific business scenarios on a regular or real-time basis, and the sources of comment feedback data include but are not limited to users' evaluations on e-commerce platforms, comments on social media, customer service conversation records, etc. By regularly obtaining users' comment feedback data, we can timely understand users' satisfaction and opinions on services and products.

[0044] It should be pointed out here that natural language processing technology (such as sentiment analysis algorithms) is used to extract sentiment data from review feedback data and identify users' positive, negative or neutral sentiments; for example, sentiment analysis tools can be used to identify users' satisfaction with recommended products or their opinions on market analysis results. At the same time, key features are extracted from review feedback data, such as user preference changes, problem feedback, and improvement suggestions.

[0045] Combine the extracted sentiment tendency data and key features with the existing user portrait data to update the target user portrait; for example, if a user gives a negative review of a recommended product, the user's preference can be adjusted in the user portrait to reduce the recommendation of similar products.

[0046] Through the above solution, by combining multi-source data and transaction scenario characteristic data, the generated preliminary user portrait is more comprehensive and accurate, and the dynamic monitoring mechanism can ensure that the user portrait can promptly reflect the latest changes in user behavior, thereby improving the real-time and effectiveness of the portrait; in addition, by regularly obtaining and analyzing user comment feedback data, the user portrait is dynamically adjusted to ensure that it always reflects the user's latest behavior and preferences, thereby improving the real-time and effectiveness of the service.

[0047] It should be pointed out here that in the process of optimizing the preliminary user portrait based on social network relationships, the method also includes: S111, based on the review feedback data, obtaining the user's evaluation data on the recommended products in different specific business scenarios and the user's opinion data on the market analysis results.

[0048] Among them, user review feedback data is collected from different business scenarios. Specific business scenarios include but are not limited to personalized recommendations, risk assessment, market analysis, etc. For example, in the personalized recommendation scenario, user evaluation data on recommended products is collected; in the market analysis scenario, user opinion data on market analysis results is collected. By collecting user evaluations and opinions in different business scenarios, we can fully understand user satisfaction and opinions on services and products.

[0049] S112, extracting the user's sentiment tendency data in the evaluation data and opinion data through a natural language processor.

[0050] Among them, natural language processing technology (such as sentiment analysis algorithm) is used to perform sentiment analysis on the collected comment feedback data, and then the sentiment tendency in the user comments is identified. It should be pointed out here that the sentiment tendency data is classified to generate sentiment labels, such as "positive", "negative", and "neutral"; and the intensity of the sentiment is evaluated, for example, the intensity of positive sentiment or the severity of negative sentiment.

[0051] Through sentiment analysis, we can accurately identify users’ emotional tendencies towards recommended products and market analysis results, understand users’ actual feelings, and quantify users’ emotional tendencies to provide specific data support for subsequent user portrait optimization.

[0052] S113, optimizing the preliminary user portrait based on the sentiment tendency data and social network relationships.

[0053] The extracted sentiment tendency data is integrated with the user's social network relationship data to form comprehensive user feature data; for example, the user's friend list, interaction records and other social network relationship data are combined with the user's sentiment tendency data for the recommended product. Based on the integrated data, corresponding feature engineering is performed to extract key features, such as the user's sentiment tendency, social influence, interaction frequency, etc. The preliminary user portrait is optimized based on the extracted key features. For example, if a user gives a negative evaluation of a recommended product and has a high influence in the social network, a "sensitive user" label can be added to the user portrait, and the recommendation strategy can be adjusted to reduce the recommendation of similar products.

[0054] By executing steps S111 to S113, by collecting the evaluations and opinions of users in different business scenarios, we can fully understand the user's satisfaction and opinions on services and products; at the same time, through sentiment analysis, we can accurately identify the user's sentiment inclination towards recommended products and market analysis results, and understand the user's actual feelings. We can also more accurately reflect the user's actual feelings and social behaviors through sentiment inclination data and social network relationship data, making the user portrait more accurate and comprehensive.

[0055] As for the system, it is necessary to have a certain early warning capability for users based on their behavior and transaction status, so the method also includes: S114, using time series analysis technology to obtain the time change trend of the user's transaction data.

[0056] Among them, the user's transaction data is obtained from multiple data sources (such as e-commerce platforms, mobile payment platforms, and bank transaction records), and then the transaction data is pre-processed, including data cleaning, format conversion, and outlier detection, to ensure the integrity and accuracy of the data. The transaction data is analyzed using time series analysis techniques (such as ARIMA, LSTM, etc.) to extract the user's time change trend, including transaction frequency, transaction amount, transaction time, login frequency, etc.

[0057] Through time series analysis, we can identify the temporal trends of user behavior and understand the user's transaction patterns and behavioral patterns. On the one hand, we can identify whether the user has periodic behavior, such as shopping habits at fixed times every month; on the other hand, we can identify the user's long-term behavioral trends, such as the gradual increase or decrease in transaction amounts.

[0058] S115, judging whether there is periodic behavior or long-term continuous trend based on the time change trend.

[0059] Among them, periodic behavior recognition: use statistical methods (such as Fourier transform, autocorrelation function) to identify periodic components in transaction data and determine whether users have fixed periodic behaviors, such as weekly shopping, monthly bill payments, etc.

[0060] Long-term trend analysis: Use trend analysis methods (such as linear regression and moving average) to identify long-term trends in transaction data and determine whether there is a long-term upward or downward trend in the user's transaction behavior, such as an increase or decrease in transaction amount year by year.

[0061] Through periodic behavior identification and long-term trend analysis, we can have a more comprehensive understanding of users' trading behavior patterns. Then, by identifying users' periodic behavior and long-term trends, it can help predict future trading behavior and provide a basis for risk management. We can also understand users' periodic behavior and long-term trends and provide more personalized services, such as regularly pushing promotional information, reminding bill payments, etc.

[0062] S116: If yes, set the corresponding threshold trigger, and generate warning data if the transaction data exceeds the threshold trigger.

[0063] Among them, reasonable threshold triggers are set according to the identified periodic behaviors and long-term trends. For example, if the user's monthly transaction amount usually fluctuates within a certain range, a threshold can be set to trigger an alarm when the transaction amount exceeds this range.

[0064] Use a real-time monitoring system (such as a streaming computing framework) to continuously monitor users' transaction data. When the transaction data exceeds the set threshold trigger, generate warning data immediately and send the warning data to relevant systems or personnel so that timely measures can be taken. In addition, the warning data includes detailed information on abnormal transactions, such as transaction time, transaction amount, transaction type, etc.

[0065] By executing steps S114 to S116, through periodic behavior identification and long-term trend analysis, a more comprehensive understanding of the user's transaction behavior pattern can be obtained, and then the user's periodic behavior and long-term trend can be identified, which will help predict future transaction behavior; timely early warning data can prompt enterprises to take preventive measures to reduce potential credit risks and fraud risks; by timely discovering and handling abnormal transactions, the user's account security can be protected and user trust and satisfaction can be improved; through real-time monitoring and early warning, the risk management process can be optimized and the stability and security of the business can be improved.

[0066] For the transaction data in this application, the corresponding data quality needs to be evaluated, so the method also includes: S101, retrieve the pre-trained quality scoring model and pre-processed transaction data.

[0067] Among them, the quality scoring model is trained using historical transaction data, which can evaluate the quality of transaction data with high accuracy and reliability. The specific training process is consistent with the modeling model and will not be repeated here. In addition, the user's transaction data is obtained from multiple data sources (such as e-commerce platforms, mobile payment platforms, and bank transaction records), and the transaction data is pre-processed, including data cleaning, format conversion, and outlier detection, to ensure the integrity and accuracy of the data.

[0068] S102, inputting the preprocessed transaction data into a quality scoring model for quality scoring, and obtaining corresponding scoring data.

[0069] Among them, the preprocessed transaction data is input into the quality scoring model, and the quality scoring model evaluates the input transaction data to generate corresponding scoring data. It should be pointed out here that the scoring data is usually a numerical value that indicates the quality level of the data, such as a score between 0 and 1, where 1 indicates the highest data quality. The quality of transaction data can be objectively evaluated through the quality scoring model, and the scoring data can help identify problems in the data, such as missing values, outliers, etc.

[0070] S103, comparing the scoring data with a preset scoring threshold, and if the scoring data is less than the scoring threshold, reacquiring corresponding transaction data.

[0071] A scoring threshold is pre-set, and the threshold is used to determine whether the data quality meets the standard. For example, the scoring threshold can be set to 0.8, indicating that only data with a score greater than or equal to 0.8 is considered qualified.

[0072] The scoring data generated by the quality scoring model is then compared with the pre-set scoring threshold. If the scoring data is less than the scoring threshold, it indicates that the data quality does not meet the standard. In this case, the corresponding transaction data is re-acquired, and the re-acquired data needs to be pre-processed and quality-scored again until the data quality meets the standard. By setting the scoring threshold and data re-acquisition mechanism, the quality of the transaction data used in the end is ensured to meet the standard, and the reliability and accuracy of the data are improved.

[0073] By executing steps S101 to S103, the quality scoring model can be used to objectively evaluate the quality of transaction data and ensure the reliability and accuracy of the data; by setting scoring thresholds and data re-acquisition mechanisms, the quality of transaction data used in the end can be ensured to meet the standards and improve the reliability and accuracy of the data. If the data quality does not meet the standards, the data can be re-acquired in a timely manner to correct the existing problems and avoid analysis errors and decision-making errors caused by data quality problems.

[0074] In the trading platform, the platform or system will recommend relevant content to the user in combination with the user's image. In order to continue to recommend to the user more accurately, the method also includes: S121, obtaining real-time behavior data of the user in real time, and obtaining context information corresponding to the real-time behavior data.

[0075] Among them, real-time data collection technology (such as log records, API interfaces, and streaming computing frameworks) is used to obtain real-time user behavior data in real time, and real-time behavior data includes but is not limited to browsing records, search records, click records, purchase records, etc. At the same time, contextual information corresponding to real-time behavior data is collected, including but not limited to time, location, device type, network environment, etc. For example, it records what operations the user performed using what device at what time and where, and what the network environment was like at the time.

[0076] S122, adjusting corresponding recommended content based on the real-time behavior data and context information.

[0077] Among them, real-time behavior data and contextual information are integrated to form a complete user behavior record. For example, the user's browsing history is combined with information such as the time, location, and device type. Then, recommendation algorithms (such as collaborative filtering and deep learning models) are used to adjust the recommended content based on the integrated data. For example, if a user browses sports shoes on a mobile phone at night, popular sports shoe brands and styles can be recommended first. According to the user's real-time behavior and contextual information, the recommended content is dynamically adjusted to ensure the accuracy and timeliness of the recommendation.

[0078] S123, obtaining corresponding recommendation feedback data based on the recommended content, and adjusting the recommendation strategy in combination with the target user portrait and the recommendation feedback data.

[0079] Among them, user feedback data on recommended content is collected, including but not limited to click-through rate, conversion rate, user evaluation, etc. For example, whether the user clicks on the recommended product, whether the recommended product is purchased, and the user's satisfaction with the recommended content is recorded. Then the feedback data is analyzed to evaluate the recommendation effect, for example, the changes in click-through rate and conversion rate are analyzed to identify the effectiveness of the recommended content and room for improvement. At the same time, the recommendation strategy is adjusted in combination with the target user portrait and feedback data. For example, if the click-through rate of a certain type of recommended content is low, the user's interest tag can be adjusted in the user portrait to reduce the recommendation of similar content.

[0080] By executing steps S121 to S123 and collecting context information, the user's real-time behavior background can be more fully understood, and the recommended content can be dynamically adjusted based on the user's real-time behavior and context information. By collecting and analyzing recommendation feedback data, forming a feedback loop, and continuously optimizing the recommendation strategy, the accuracy and conversion rate of the recommendation can be improved; by combining the target user portrait and feedback data, more personalized recommendation content can be provided to improve user experience and user satisfaction.

[0081] In addition, in order to adjust the corresponding recommendation strategy in time, the method also includes: S124, obtaining the user's corresponding interaction record data.

[0082] Among them, the user's previous interaction record data is obtained from channels such as the customer service system, complaint management system, and service request system, and the interaction record data includes but is not limited to customer service conversation records, complaint records, service request records, etc. The collected interaction record data is then sorted to ensure the structuring and standardization of the data for subsequent processing. For example, the text content in the customer service conversation record is cleaned and formatted, and the key information in the complaint record and service request record is extracted.

[0083] S125, identifying the preference data and demand data corresponding to the user based on the previous interaction record data.

[0084] Among them, natural language processing technology (such as text classification and sentiment analysis) is used to parse customer service conversation records and identify user needs and preferences. For example, sentiment analysis is used to identify the user's emotions in the conversation, and entity recognition is used to extract the products or services mentioned by the user. Common user problems and needs can be identified by analyzing complaint records and service request records. For example, keywords in complaint records can be analyzed to identify user dissatisfaction with product quality or service attitude; the content in service request records can be analyzed to identify the user's specific needs. In addition, user preference data, such as favorite brands, commonly used service types, etc., are extracted from previous interaction records. For example, brand names mentioned by users many times can be analyzed to identify user brand preferences.

[0085] S126, adjusting the recommendation strategy based on the preference data and the demand data, and recommending the latest recommended content to the user based on the adjusted recommendation strategy.

[0086] The recommendation strategy can be adjusted based on the user's demand data and preference data. For example, if users often complain about product quality issues, the recommendation strategy can reduce the recommendation of similar products; if users prefer a certain brand, the recommendation of that brand can be increased in the recommendation strategy.

[0087] As mentioned above, recommendation algorithms (such as collaborative filtering and deep learning models) can be used to generate the latest recommendation content. For example, based on the adjusted recommendation strategy, the products or services that the user is most likely to be interested in are generated. The generated recommendation content can be pushed to the user through various channels (such as APP notifications, emails, and text messages). For example, the latest promotions can be recommended to users through APP notifications, and personalized products can be recommended to users through emails.

[0088] By executing steps S124 to S126 and collecting the user's previous interaction records, we can fully understand the user's historical behavior and needs, so that we can accurately identify the user's actual needs and preferences, and identify the problems encountered by the user during use, solve the user's problems in a timely manner, and improve user satisfaction.

[0089] In addition, based on user needs and preferences, more personalized recommendations can be provided; and by optimizing recommendation strategies, the accuracy and conversion rate of recommendations can be improved.

[0090] Finally, for trading platforms, competitors in the market will seize user resources, so the platform also needs to conduct certain user monitoring. In order to reduce the problem of user loss, this method also includes: S127, retrieve key features corresponding to user-related data.

[0091] Among them, relevant user data is collected from multiple data sources (such as e-commerce platforms, mobile payment platforms, bank transaction records, customer service systems, etc.). Key features include but are not limited to transaction frequency, transaction amount, transaction time, login frequency, number of interactions, number of complaints, customer service conversation records, etc.

[0092] S128, determining the churn level corresponding to the user based on the key features in combination with the time series.

[0093] Among them, time series analysis techniques (such as ARIMA, LSTM, etc.) are used to analyze key features and identify the time-varying trends of user behaviors, for example, analyzing the changing trends of users' transaction frequency, transaction amount, login frequency, etc. over time. Then, features related to user churn are extracted from the time series analysis results, for example, identifying the downward trend of user transaction frequency, the decrease in login frequency, the increase in the number of complaints, etc.

[0094] Based on the extracted churn features, machine learning algorithms (such as decision trees, random forests, logistic regression, etc.) are used to determine the user's churn level. For example, users are divided into three levels: high churn risk, medium churn risk, and low churn risk.

[0095] Through time series analysis, we can identify the time-varying trends of user behavior, understand user behavior patterns, accurately predict user churn risks, and determine user churn levels, so that we can promptly identify users with high churn risks, take preventive measures, and reduce user churn rates.

[0096] S129, generating corresponding retention measures based on the churn level.

[0097] Among them, according to the user's churn level, design corresponding retention measures, including but not limited to personalized discounts, customized services, proactive contact and satisfaction surveys.

[0098] Specifically, personalized offers: For users with a high risk of churn, personalized offers are provided, such as discount coupons, membership benefits, etc. For example, exclusive discount codes are sent to users with a high risk of churn to attract them to continue using the service. Customized services: For users with a medium risk of churn, customized services are provided, such as exclusive customer service, customized product recommendations, etc. For example, one-on-one customer service support is provided to users with a medium risk of churn to answer their questions and provide personalized service suggestions. Proactive contact: For users with a low risk of churn, regular proactive contact is made to understand their needs and opinions. For example, regular communication with users is conducted through telephone, email, etc. to provide care and support. Satisfaction surveys: Regular satisfaction surveys are conducted on all users to collect user feedback and suggestions. For example, through online questionnaires or telephone surveys, users’ satisfaction with the service is understood and services are improved in a timely manner.

[0099] By executing steps S127 to S129, user problems can be discovered and solved in a timely manner, thereby improving user trust and satisfaction with the service; and through effective retention measures, the user churn rate can be reduced, the user retention rate can be increased, and the business results and market competitiveness can be improved.

[0100] The present application embodiment discloses a user big data portrait generation device based on a transaction scenario, referring to Figure 2 , the devices include but are not limited to: The transaction scenario characteristic data acquisition module 200 acquires the transaction data of the user from multiple different data sources, and pre-processes the acquired transaction data, and acquires the corresponding transaction scenario characteristic data based on the data source, and combines the pre-processed transaction data with the corresponding transaction scenario characteristic data to perform modeling using a pre-trained modeling model to generate a corresponding preliminary user portrait, wherein the data source includes an e-commerce platform, a mobile payment platform, and a bank transaction record; The target user portrait acquisition module 210 applies the generated user portrait to specific business scenarios, obtains the social network relationships of users in different specific business scenarios, and optimizes the preliminary user portrait based on the social network relationships to obtain the corresponding target user portrait, wherein the specific business scenarios include personalized recommendations, risk assessment, and market analysis; The target user portrait updating module 220 calls a pre-established dynamic monitoring mechanism for user behavior, and regularly obtains user comment feedback data corresponding to specific business scenarios based on the dynamic monitoring mechanism, and regularly updates the corresponding target user portrait based on the comment feedback data.

[0101] Furthermore, the device includes but is not limited to: The opinion data acquisition module is used to obtain users' evaluation data on recommended products in different specific business scenarios and their opinions on market analysis results based on review feedback data; The sentiment tendency data extraction module is used to extract the sentiment tendency data of users in the evaluation data and opinion data through the natural language processor; The preliminary user portrait optimization module is used to optimize the preliminary user portrait based on sentiment tendency data and social network relationships.

[0102] Furthermore, the device includes but is not limited to: The time change trend acquisition module uses time series analysis technology to obtain the time change trend of the user's transaction data, where the time change trend includes transaction frequency, transaction amount, transaction time, and login frequency; The judgment module is used to judge whether there is periodic behavior or long-term continuous trend based on the time change trend; The warning data generation module, if any, sets the corresponding threshold trigger, and is used to generate warning data if the transaction data exceeds the threshold trigger.

[0103] Furthermore, the device includes but is not limited to: The quality scoring model retrieval module is used to retrieve the pre-trained quality scoring model and pre-processed transaction data; The scoring data acquisition module inputs the pre-processed transaction data into the quality scoring model for quality scoring, so as to obtain the corresponding scoring data; The comparison module is used to compare the scoring data with a preset scoring threshold. If the scoring data is less than the scoring threshold, the corresponding transaction data is re-acquired.

[0104] Furthermore, the device includes but is not limited to: A real-time behavior data acquisition module is used to acquire the user's real-time behavior data and the context information corresponding to the real-time behavior data. The real-time behavior data includes but is not limited to browsing history, search history, click history, and purchase history. The context information includes but is not limited to time, location, device type, and network environment. A recommended content adjustment module, which is used to adjust the corresponding recommended content based on real-time behavior data and context information; The recommendation feedback data acquisition module is used to obtain the corresponding recommendation feedback data based on the recommended content, and adjust the recommendation strategy based on the target user portrait and the recommendation feedback data.

[0105] Furthermore, the device includes but is not limited to: The module for acquiring the interaction record data is used to acquire the interaction record data corresponding to the user, wherein the interaction record data includes customer service conversation records, complaint records, and service request records; A preference data identification module is used to identify the user's corresponding preference data and demand data based on the previous interaction record data; The recommendation strategy adjustment module is used to adjust the recommendation strategy based on the preference data and the demand data, and recommend the latest recommended content to the user based on the adjusted recommendation strategy.

[0106] Furthermore, the device includes but is not limited to: Key feature retrieval module, used to retrieve key features corresponding to user-related data, including transaction frequency, transaction amount, transaction time, login frequency, number of interactions, number of complaints, and customer service conversation records; The churn level determination module is used to determine the churn level corresponding to the user based on the key features in combination with the time series; The retention measures generation module is used to generate corresponding retention measures based on the churn level, wherein the retention measures include personalized discounts, customized services, proactive contact and satisfaction surveys.

[0107] An embodiment of the present application also discloses a user big data portrait generation system based on transaction scenarios, including a processor, in which a program of any one of the above-mentioned user big data portrait generation methods based on transaction scenarios runs.

[0108] An embodiment of the present application also discloses a storage medium storing a program of any one of the above-mentioned methods for generating user big data portraits based on transaction scenarios.

[0109] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for generating user big data portraits based on transaction scenarios, characterized in that: include: Acquire the user's transaction data from multiple different data sources, pre-process the acquired transaction data, acquire corresponding transaction scenario characteristic data based on the data source, combine the pre-processed transaction data with the corresponding transaction scenario characteristic data using a pre-trained modeling model to perform modeling, and generate a corresponding preliminary user portrait, wherein the data source includes an e-commerce platform, a mobile payment platform, and a bank transaction record; Apply the generated user portrait to specific business scenarios, obtain the social network relationships of users in different specific business scenarios, optimize the preliminary user portrait based on the social network relationships, and obtain the corresponding target user portrait, wherein the specific business scenarios include personalized recommendations, risk assessment, and market analysis; A pre-established dynamic monitoring mechanism for user behavior is retrieved, and comment feedback data corresponding to users in specific business scenarios is regularly obtained based on the dynamic monitoring mechanism. Based on the comment feedback data, the corresponding target user portrait is regularly updated.

2. The method for generating user big data portrait based on transaction scenarios according to claim 1 is characterized in that: In the process of optimizing the preliminary user portrait based on the social network relationship, the method further includes: Based on the review feedback data, obtain the user's evaluation data on the recommended products in different specific business scenarios and the opinion data on the market analysis results; Extracting the user's sentiment tendency data in the evaluation data and the opinion data by a natural language processor; The preliminary user portrait is optimized based on the sentiment tendency data and the social network relationship.

3. The method for generating user big data portrait based on transaction scenarios according to claim 2 is characterized in that: The method also includes: Use time series analysis technology to obtain the time change trend of the user's transaction data, where the time change trend includes transaction frequency, transaction amount, transaction time, and login frequency; Determine whether there is a periodic behavior or a long-term continuous trend based on the time change trend; If so, set the corresponding threshold trigger. If the transaction data exceeds the threshold trigger, generate warning data.

4. The method for generating user big data portrait based on transaction scenarios according to claim 3 is characterized in that: The method also includes: Retrieve the pre-trained quality scoring model and pre-processed transaction data; Inputting the preprocessed transaction data into a quality scoring model for quality scoring, and obtaining corresponding scoring data; The scoring data is compared with a preset scoring threshold, and if the scoring data is less than the scoring threshold, the corresponding transaction data is reacquired.

5. The method for generating user big data portrait based on transaction scenarios according to claim 1 is characterized in that: The method also includes: Acquire the user's real-time behavior data in real time, and acquire the context information corresponding to the real-time behavior data, the real-time behavior data including but not limited to browsing history, search history, click history, purchase history, the context information including but not limited to time, location, device type, network environment; Adjusting corresponding recommended content based on the real-time behavior data and the context information; Based on the recommended content, corresponding recommendation feedback data is obtained, and the recommendation strategy is adjusted in combination with the target user portrait and the recommendation feedback data.

6. The method for generating user big data portrait based on transaction scenarios according to claim 5 is characterized in that: The method also includes: Obtain the user's corresponding interaction record data, where the interaction record data includes customer service conversation records, complaint records, and service request records; Identify the preference data and demand data corresponding to the user based on the previous interaction record data; The recommendation strategy is adjusted based on the preference data and the demand data, and the latest recommended content is recommended to the user based on the adjusted recommendation strategy.

7. The method for generating user big data portrait based on transaction scenarios according to claim 6 is characterized in that: The method also includes: Retrieve key features corresponding to user-related data, including transaction frequency, transaction amount, transaction time, login frequency, number of interactions, number of complaints, and customer service conversation records; Determine the churn level corresponding to the user based on the key features in combination with the time series; Corresponding retention measures are generated based on the churn level, wherein the retention measures include personalized discounts, customized services, proactive contact, and satisfaction surveys.

8. A user big data portrait generation device based on transaction scenarios, characterized in that: include: A transaction scenario characteristic data acquisition module acquires user transaction data from multiple different data sources, and pre-processes the acquired transaction data, and acquires corresponding transaction scenario characteristic data based on the data source, and combines the pre-processed transaction data with the corresponding transaction scenario characteristic data to perform modeling using a pre-trained modeling model to generate a corresponding preliminary user portrait, wherein the data source includes an e-commerce platform, a mobile payment platform, and bank transaction records; A target user portrait acquisition module applies the generated user portrait to specific business scenarios, obtains the social network relationships of users in different specific business scenarios, and optimizes the preliminary user portrait based on the social network relationships to obtain the corresponding target user portrait, wherein the specific business scenarios include personalized recommendations, risk assessment, and market analysis; The target user portrait update module calls a pre-established dynamic monitoring mechanism for user behavior, and regularly obtains user comment feedback data corresponding to specific business scenarios based on the dynamic monitoring mechanism, and regularly updates the corresponding target user portrait based on the comment feedback data.

9. A user big data portrait generation system based on transaction scenarios, characterized in that: It includes a processor, in which runs a program of a method for generating a user big data portrait based on a transaction scenario as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: A program storing a method for generating a user big data portrait based on a transaction scenario as described in any one of claims 1 to 7.

Images