Credit marketing method and device based on user portrait and electronic equipment

Abstract

The invention provides a credit marketing method and device based on a user portrait and electronic equipment, and the method comprises the steps: obtaining the user data of a stock user, and carrying out the feature extraction of the user data, and obtaining a user feature vector; wherein the user data at least comprises user basic information, behavior data, credit data and user interest data; generating a user portrait of each stock user based on the user feature vector and a pre-constructed user portrait model; based on the user portrait and a preset user screening condition for the target product, screening the stock users to obtain a target user; generating marketing content for the target user based on the user portrait of the target user and the marketing activity of the target product; and determining a marketing channel based on the user portrait of the target user, and pushing the marketing content to the corresponding target user based on the marketing channel. According to the invention, the marketing effect and accuracy of credit products are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of financial credit, and particularly to a credit marketing method, device and electronic device based on user portraits. Background Art

[0002] With the continuous development of the financial market and the increasing competition, credit institutions are facing greater challenges in customer management and marketing. The traditional credit customer operation methods mainly rely on manual operations, which are inefficient and difficult to accurately reach target customers. At the same time, with the rapid development of technologies such as big data and artificial intelligence, automated marketing has become a key means for the financial industry to enhance competitiveness. However, existing automated marketing systems often lack in-depth understanding of the characteristics of existing credit customers and targeted operation strategies, and it is difficult to meet the actual needs of credit institutions. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide a credit marketing method, device and electronic device based on user portraits, so as to improve the marketing effect and accuracy of credit products.

[0004] To achieve the above purpose, the technical solutions adopted by the present invention are as follows:

[0005] In the first aspect, the present invention provides a credit marketing method based on user portraits, including: obtaining user data of existing users, and extracting features from the user data to obtain user feature vectors; where the user data at least includes user basic information, behavior data, credit data, and user interest data; generating user portraits of each existing user based on the user feature vectors and a pre-constructed user portrait model; screening the existing users based on the user portraits and pre-set user screening conditions for the target product to obtain target users; generating marketing content for the target users based on the user portraits of the target users and the marketing activities of the target product; determining a marketing channel based on the user portraits of the target users, and pushing the marketing content to the corresponding target users through the marketing channel.

[0006] Optionally, extracting features from the user data to obtain user feature vectors includes: extracting initial user features from the user data; screening the initial user features based on a preset feature selection algorithm to obtain user features, and encoding the user features to obtain user feature vectors; where the preset features include one of the following: filter method, wrapper method, and embedding method.

[0007] Optionally, user portraits of each existing user are generated based on the user feature vector and the pre-constructed user portrait model, including: constructing a user portrait hierarchical architecture based on the product requirements of the target product; constructing a combined feature vector for each level of the user portrait based on the user feature vector and the pre-constructed user portrait model; calculating the information entropy of the combined feature vector and the user feature vector, and determining the user features of each level of the user portrait based on the information entropy to generate the user portrait of each level.

[0008] Optionally, marketing content is pushed to the corresponding target users based on the marketing channels, including: classifying the target users based on their basic information to obtain the targeted designated users and the excluded users, and pushing the marketing content to the targeted designated users; determining the secondary push users based on the business data of the target users, and pushing the marketing content to the secondary push users.

[0009] Optionally, after pushing the marketing content to the corresponding target users based on the marketing channels, it further includes: obtaining the execution information of the marketing activity; where the execution information includes one or more of the following: the sending time of the marketing content, the sent content, and the sending channel; obtaining the feedback information of the target users on the marketing activity; where the feedback information includes one or more of the following: click-through rate, conversion rate, and return on investment; optimizing the marketing activity and the user portrait based on the execution information and the feedback information.

[0010] Optionally, marketing content for the target users is generated based on the user portraits of the target users and the marketing activities of the target product, including: constructing prompt words for the large language model according to the user portraits of the target users and the marketing activities of the target product; inputting the prompt words into the large language model to generate the marketing content for the target users.

[0011] Optionally, the construction of the user portrait model includes: obtaining the user data of the user, and determining the portrait labels of the user data based on the pre-constructed multi-dimensional label system; where the label system includes at least: behavior labels, interest labels; obtaining the real-time behavior data of the user, and updating the portrait labels based on the real-time behavior data; training the machine learning model based on the user data and the portrait labels to obtain the user portrait model.

[0012] In a second aspect, the present invention provides a credit marketing device based on user portraits, comprising: a feature extraction module for obtaining user data of existing users and extracting features from the user data to obtain user feature vectors; wherein the user data includes at least user basic information, behavior data, credit data, and user interest data; a user portrait generation module for generating user portraits of each existing user based on the user feature vectors and a pre-constructed user portrait model; a target user screening module for screening existing users based on the user portraits and pre-set user screening conditions for target products to obtain target users; a marketing content generation module for generating marketing content for target users based on the user portraits of the target users and marketing activities of the target products; and a marketing content pushing module for determining marketing channels based on the user portraits of the target users and pushing the marketing content to the corresponding target users through the marketing channels.

[0013] In a third aspect, the present invention provides an electronic device comprising a processor and a memory, the memory storing computer-executable instructions capable of being executed by the processor, and the processor executing the computer-executable instructions to implement the steps of any one of the methods provided in the first aspect above.

[0014] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the steps of any one of the methods provided in the first aspect above.

[0015] The present invention brings the following beneficial effects:

[0016] The above-mentioned credit marketing method, device and electronic device based on user portraits provided by the present invention first obtain user data of existing users (including at least user basic information, behavior data, credit data, and user interest data), and extract features from the user data to obtain user feature vectors; secondly, generate user portraits of each existing user based on the user feature vectors and a pre-constructed user portrait model; then screen existing users based on the user portraits and pre-set user screening conditions for target products to obtain target users; then generate marketing content for target users based on the user portraits of the target users and marketing activities of the target products; and finally determine marketing channels based on the user portraits of the target users and push the marketing content to the corresponding target users through the marketing channels. In the above method, user portraits can be constructed based on multi-dimensional user data, improving the accuracy of user portraits; at the same time, existing users are screened according to the user portraits and user screening conditions for target products to obtain target users, and then personalized marketing content for each target user is generated and the marketing content is pushed to the corresponding target users through marketing channels, thereby improving the marketing effect and accuracy of credit products.

[0017] Other features and advantages of the present invention will be set forth in the following description, and in part will be obvious from the description, or may be learned by practice of the present invention. The objectives and other advantages of the present invention are realized and attained by the structure particularly pointed out in the specification, claims and drawings.

[0018] In order to make the above objectives, features and advantages of the present invention more obvious and understandable, the following specifically gives preferred embodiments in conjunction with the accompanying drawings and describes them in detail as follows. Brief Description of the Drawings

[0019] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0020] Figure 1 It is a flowchart of a credit marketing method based on user portraits provided by an embodiment of the present invention;

[0021] Figure 2 It is a flowchart of another credit marketing method based on user portraits provided by an embodiment of the present invention;

[0022] Figure 3 It is a schematic structural diagram of a credit marketing device based on user portraits provided by an embodiment of the present invention;

[0023] Figure 4 It is a schematic structural diagram of an electronic device provided by an embodiment of the present invention. Detailed Embodiments

[0024] In order to make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions of the present invention in conjunction with the drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0025] Currently, existing automated marketing systems often lack in-depth understanding of the characteristics of existing credit customers and targeted operation strategies, and it is difficult to meet the actual needs of credit institutions. There are still some deficiencies and defects, which are mainly reflected in the following aspects:

[0026] (1) User data is scattered in various financial institutions, Internet platforms, etc., lacking an effective integration and sharing mechanism, resulting in an incomplete and inaccurate user portrait.

[0027] (2) Most of the existing user portrait models are based on static data and simple rules, making it difficult to capture the dynamic changes and complex relationships of user behaviors, resulting in inaccurate user portraits and poor marketing effects.

[0028] (3) Most of the existing systems adopt fixed marketing strategies, lacking personalization and intelligence, and it is difficult to meet the differentiated needs of different users, thus resulting in poor user experience and low marketing conversion rate.

[0029] (4) The construction and application of user portraits involve a large amount of user privacy data, posing risks of data leakage and abuse.

[0030] (5) Most of the existing systems adopt complex algorithm models, lacking interpretability and transparency, and it is difficult to understand the generation process of user portraits and the basis for formulating marketing strategies.

[0031] (6) Building and deploying a credit marketing system based on user portraits requires a large amount of manpower, material resources and financial resources, which is costly for small and medium-sized financial institutions.

[0032] Based on this, a credit marketing method, device and electronic device provided by an embodiment of the present invention can improve the marketing effect and accuracy of credit products.

[0033] For the convenience of understanding this embodiment, first, a credit marketing method based on user portraits disclosed in an embodiment of the present invention will be introduced in detail. This method can be executed by an electronic device, such as a smart phone, a computer, a tablet, etc. Refer to Figure 1 The flowchart of a credit marketing method based on user portraits shown, which shows that the device mainly includes the following steps S101 to step S105:

[0034] Step S101: Obtain user data of existing users, and perform feature extraction on the user data to obtain user feature vectors.

[0035] In one implementation, user data can be collected from various business systems (such as: CRM system, credit system, website, APP, etc.) through API interfaces, data file imports, etc. Specifically, in this embodiment, data collection supports multiple data access methods, including network APIs, email attachments, FTP / SFTP transfers, etc., to ensure that data can be collected from different customers and outbound call channels, and then integrate data from different sources, create a unified data view, and apply the ETL (Extract, Transform, Load) process to extract, transform and load the data. The import of data sources supports structured and unstructured data, mainly dealing with a large amount of unstructured data inside and outside financial institutions, and batch processing and analysis of the data stored therein.

[0036] User data includes: basic user information, behavioral data, credit data, user interest data, other data (social data, consumption data), etc. Among them, basic user information includes age, gender, region, occupation, income, operator, mobile phone number location, etc.; behavioral data includes website browsing records, APP usage records, search records, click records, etc.; credit data includes loan records, repayment records, credit card usage records, etc.; user interest data includes travel, food, technology, etc.

[0037] Preprocess the collected data, including cleaning, deduplication, and completion operations, such as processing missing values, outliers, and duplicate data, unifying data formats and units, removing duplicate data, verifying data validity, and ensuring data quality. You can also use tools such as Dagster and dbt to orchestrate data processing tasks to ensure automation and repeatability of data processing.

[0038] The cleaned data is stored in a database or data warehouse, and a unique user ID is established to achieve unified management of user data. Specifically, in data storage, a hierarchical storage strategy can be adopted according to the timeliness of the data to store hot data, warm data, and cold data in different database systems. At the same time, relational databases (such as PostgreSQL) and non-relational databases (such as MongoDB) are used to meet different storage needs.

[0039] Furthermore, feature extraction is performed on the cleaned user data to screen out the most useful features for constructing the user portrait, specifically including: firstly, feature extraction is performed on the user data to obtain initial user features; then, the initial user features are screened based on a preset feature selection algorithm to obtain user features, and the user features are encoded to obtain a user feature vector; wherein the preset features include one of the following: filtering method, packaging method and embedding method.

[0040] In the specific implementation, firstly, meaningful initial user features are extracted from the original user data, for example, preference categories are extracted from the user's browsing history; then, a suitable feature selection algorithm is selected for feature selection according to the data characteristics and task requirements; finally, the selected user features are encoded to obtain the user feature vector. Common feature selection methods include:

[0041] Filtering: Feature selection based on statistical tests or scoring mechanisms independent of any machine learning algorithm. For example, chi-square test, mutual information, etc.

[0042] Wrapping method: Run multiple rounds of training with a predictive model, each round using a different subset of features, and select the best feature set based on model performance.

[0043] Embedded method: Feature selection is automatically performed during model training. LASSO and Ridge regression are two common embedded methods that achieve feature selection by imposing penalties on feature coefficients.

[0044] Step S102: Generate a user profile for each existing user based on the user feature vector and the pre-constructed user profile model.

[0045] In one implementation, machine learning algorithms (such as logistic regression, decision tree, random forest, neural network, etc.) can be used to train the user profile model based on historical user data and the user feature vector. The trained user profile model can predict user demographic information, hobbies, consumption ability, risk preference, etc.

[0046] Based on this, in the embodiments of the present invention, the extracted user feature vector can be input into the pre-constructed user profile model to profile the existing users and generate a user profile for each existing user, including: user attribute tags (age, gender, region, etc.), interest tags (travel, food, technology, etc.), consumption ability tags (high consumption, medium consumption, low consumption, etc.), risk tags (conservative, stable, aggressive, etc.), etc. As user data accumulates continuously and business requirements change, the user profile can be updated regularly to ensure the accuracy and timeliness of the profile.

[0047] Step S103: Screen the existing users based on the user profile and the pre-set user screening conditions for the target product to obtain the target users.

[0048] In one implementation, the pre-set user screening conditions for the target product can include age, income, credit score, etc. Such as: 1) Age range: 25 - 35 years old; 2) Income level: monthly income above 10,000 yuan; 3) Credit score: above 700 points.

[0049] In this embodiment, potential customers who meet the conditions, that is, target users, can be screened out from the existing users according to the user profile and the pre-set user screening conditions for the target product, and a target user list can be generated.

[0050] In some embodiments, tools and interfaces can also be provided for business personnel to enable them to generate target users according to the pre-set marketing strategies. In addition, complex population screening logics (i.e., user screening conditions) are also supported, including factors such as population attributes, conversion situations, outbound call intentions, historical marketing records, etc.

[0051] Step S104: Generate marketing content for the target users based on the user profiles of the target users and the marketing activities of the target product.

[0052] In one implementation, personalized marketing content for target users can be automatically generated based on the target customer profile and the marketing activities of the target product (i.e., the characteristics of the credit product). For example: (1) Loan amount: The maximum loan amount can be up to 500,000 yuan. (2) Interest rate: The annual interest rate is as low as 5%. (3) Repayment method: Equal principal and interest, equal principal, etc. Specifically, natural language generation technology can be used to generate the copywriting of personalized marketing content, such as email content, text message content, etc.

[0053] Step S105: Determine the marketing channels based on the user profile of the target user, and push the marketing content to the corresponding target users based on the marketing channels.

[0054] In one implementation, the appropriate marketing channels can be selected according to the user profile and behavior habits of the target customer, such as text messages, emails, APP push, etc. Then, automated marketing is carried out through channels such as intelligent outbound call robots and text messages, and the marketing content is pushed to the target customers. Among them, (1) Text messages: Suitable for sending concise and clear marketing information. (2) Emails: Suitable for sending detailed marketing content. (3) APP push: Suitable for sending personalized marketing information.

[0055] The above-mentioned credit marketing method based on user profile provided by the present invention can construct user profiles according to multi-dimensional user data, improving the accuracy of user profiles; at the same time, the existing users are screened according to the user profile and the user screening conditions for the target product to obtain the target users, and then personalized marketing content for each target user is generated, and the marketing content is pushed to the corresponding target users through the marketing channels, thereby improving the marketing effect and accuracy of the credit product.

[0056] In one implementation, in order to improve the accuracy of the user profile, a hierarchical construction method of the user profile is adopted in the implementation of the present invention. Based on this, for the aforementioned step S102, that is, when generating the user profile of each existing user based on the user feature vector and the pre-constructed user profile model, the following methods can be adopted, including but not limited to: First, construct a hierarchical architecture of the user profile based on the product requirements of the target product; then, construct the combined feature vector of each level of user profile based on the user feature vector and the pre-constructed user profile model; finally, calculate the information entropy of the combined feature vector and the user feature vector, and determine the user features of each level of user profile based on the information entropy to generate the user profile of each level.

[0057] In specific implementation, first, the user portraits need to be classified according to the portrait architecture, that is, to construct a hierarchical architecture for user portraits. Since the dimensions of the portraits are different and the portrait results vary greatly, an effective portrait must be a set of labels targeted at business results. For example, the business purpose of loan operation is the precise marketing of loan products. Therefore, the grading basis of the portrait architecture is the existing product system. A three-level user portrait architecture is established from three dimensions: the loan demand of users, service matching, and touch channels.

[0058] Among them, loan demand: This level of portrait mainly depicts the intensity of users' loan desires.

[0059] Service matching: This level of portrait depicts whether the loan amount, interest rate, and term match the users' expectations.

[0060] Touch channels: Among channels such as emails, text messages, phone calls, and official account messages, which channels are users more likely to respond to the information of.

[0061] For each level of user portrait, the user characteristics can be automatically constructed through machine learning to form a user portrait. The trees learned by the GBDT model are used to construct combined features, so that a variety of discriminative features and combined features can be discovered, which helps to automatically construct user portraits. The specific steps for constructing each level of user portrait are as follows:

[0062] Step 1: Based on the dimensions in the user data, construct the labels (label) to be learned. For example, the distinction of the strength of user needs is as follows: Users who visit the loan page on our bank's WeChat work account more than three times within one month or have completed credit are recorded as strong, and the rest are recorded as weak.

[0063] Step 2: Train the GBDT model with the labeled data obtained in the previous step, then use the trees learned by the GBDT model to construct a combined feature vector. After that, calculate the information entropy of the combined feature vector and the original user feature vector, and determine the importance of these features according to the information entropy.

[0064] Step 3: Based on the information entropy of the combined feature vector and the original user feature vector, screen out the user features, which are recorded as the user portraits at this level. The user portraits can support the marketing decisions of business personnel. For example, screening the user groups for a certain marketing activity.

[0065] Step 4: Combine the user features screened in the previous step with the existing features to train a Logistic Regression (LR) model, obtain the value scores of each user, and important users can be screened according to the value scores of the users.

[0066] In one implementation, to enhance the marketing effect, large language models can be used to generate personalized marketing content. Specifically, first, prompts for the large language model are constructed based on the user profile of the target user and the marketing activities of the target product; then the prompts are input into the large language model to generate marketing content for the target user.

[0067] In specific implementation, first, a suitable large language model is selected to generate marketing content according to the specific problems to be solved and resource constraints. Among them, the factors to be considered when selecting a large language model may include the capabilities of the model, the relevance of the training dataset, cost, and accessibility, etc. If the pre-trained model cannot directly meet the requirements, the model can be fine-tuned with data in a specific domain or task to improve its performance in personalized content generation. Then, basic templates are designed for different types of marketing content (such as emails, social media posts, product recommendations, etc.), and it is determined which parts will be dynamically filled by the LLM. Finally, the LLM is called to generate content, and the user profile and relevant parameters are input into the LLM to generate personalized marketing content for each user. This step may involve natural language processing technologies such as text summarization and question-and-answer systems. Specifically, prompts for the large language model are constructed based on the user profile of the target user and the marketing activities of the target product, and the prompts are input into the large language model to generate marketing content for the target user.

[0068] In the embodiments of the present invention, by leveraging the power of large language models, not only can a large number of high-quality personalized contents be produced more efficiently, but also the user engagement and satisfaction can be significantly improved, thus promoting business growth.

[0069] In one implementation, to meet the requirements of different marketing activities and enhance the marketing effect, when pushing marketing content to corresponding target users based on marketing channels, the following methods can be adopted, including but not limited to: First, classify the target users based on their basic information to obtain targeted designated users and excluded users, and push the marketing content to the targeted designated users; then determine secondary push users based on the business data of the target users, and push the marketing content to the secondary push users.

[0070] In specific implementation, the target users are screened according to business rules, and the population that meets specific marketing strategies is filtered out for pushing marketing content. The screening of the population specifically includes two stages:

[0071] (1) The first stage: Based on the basic information of the target users, the population is screened according to each field, and the whole or part of the population is called once.

[0072] Specifically, according to the basic information of existing users, relevant fields are used to group them, and then the target users are called as a whole or partially. Commonly used screening fields include: 1) operator; 2) group type: first loan-mob4, repeat loan-m12; 3) marketing level (level1~level10).

[0073] After the target users are screened, the target users can be divided into: targeted designated users and excluded users. Targeted designated users refer to users who need targeted push, and excluded users refer to users who do not need push.

[0074] (2) The second stage: Based on the user’s basic information and business data, the group is selected and some of them are called again.

[0075] Specifically, in the second phase of push notifications, in addition to using the user's basic information, the business data generated during the business process can also be used to select the audience. The main business data used are:

[0076] 1) Data sent back by the outbound call channel, such as call status and intent level. In actual crowd selection, some clear tags need to be processed for use, such as whether a user has been connected or reached.

[0077] 2) Analysis of user call recording data: Some labels can also be abstracted from the recording data.

[0078] It should be noted that compared with static user basic information, user business data is updated dynamically. For example, if a user was previously in an unconnected state, as the outbound call progresses, the user's state may change to connected.

[0079] In the embodiment of the present invention, the following intelligent marketing tools can be used to push marketing content:

[0080] (1) Intelligent marketing operation platform, integrating user profiling, DMP (data management platform), precision marketing, intelligent interaction and other functions.

[0081] (2) Use intelligent outbound calling robots, text messages, emails and other channels to achieve automated and large-scale marketing.

[0082] (3) Use AI voice robots to make early batch calls, explore customer intent, and intelligently allocate human agents for further marketing.

[0083] During the delivery of marketing content, manage the data consumption of outbound calling activities, including the maintenance of outbound call queues, the recording of outbound call results, etc.

[0084] (1) Provide tools and platforms for back-end technical personnel to implement complex population selection strategies and data analysis strategies.

[0085] (2) Integrated machine learning models, including algorithms such as logistic regression, decision trees, random forests, neural networks, etc., are used to predict customer behavior and provide personalized recommendations.

[0086] Build a suitable digital marketing platform in combination with actual marketing needs. Through a bottom-up customer label screening system and a marketing prediction model, corresponding marketing lists and marketing plans are formed, which cooperate with functions such as customer information, label management, target customer screening, and marketing activity management to create a marketing closed-loop and achieve precise marketing.

[0087] In one implementation, after pushing marketing content to corresponding target users based on marketing channels, the above method further includes:

[0088] First, obtain the execution information of the marketing activity.

[0089] In specific implementation, after generating marketing content, personalized marketing content can be pushed to target customers through the selected marketing channels, and the execution information of the marketing activity is recorded. Among them, the execution information includes one or more of the following: the sending time of the marketing content, the sent content, and the sending channel.

[0090] Then, obtain the feedback information of the target users on the marketing activity.

[0091] In specific implementation, the program collects the feedback information of the target users on the marketing activity, such as click-through rate, conversion rate, and return on investment, etc. Specifically, relevant data of the marketing activity can be collected according to feedback such as whether the user clicks on the marketing link, whether the user fills out a loan application, and whether the user completes a loan signing, such as marketing activity exposure, marketing activity clicks, loan application volume, loan signing volume, and then calculate the click-through rate, conversion rate, and return on investment, etc., and store the feedback information of the users in the database for subsequent analysis and optimization.

[0092] Finally, optimize the marketing activity and user portrait based on the execution information and feedback information.

[0093] In specific implementation, evaluate the effect of the marketing activity, such as calculating indicators such as click-through rate, conversion rate, and ROI (Return on Investment).

[0094] (1) Click-through rate (CTR) = number of clicks / number of exposures * 100%;

[0095] (2) Conversion rate = number of conversions / number of clicks * 100%;

[0096] (3) ROI = (revenue - cost) / cost * 100%.

[0097] Furthermore, an evaluation report on the effectiveness of marketing campaigns can be generated to help users understand the effectiveness of marketing campaigns and analyze marketing campaigns, such as outbound call data, optimize outbound call strategies, and improve outbound call efficiency and conversion rates.

[0098] The above method provided by the embodiments of the present invention can access the online channel buried point data, realize the real-time acquisition and feature analysis of customer behavior data and click stream data, and realize real-time event marketing. Through the built-in policy engine, flexible and visual configuration of marketing rules can be realized, and it can be connected to the background, associated with customer groups, and realize the rapid release and rapid adjustment of marketing campaigns. By monitoring and analyzing the operation of the model, the execution of marketing campaigns, etc., a comprehensive analysis of the effectiveness of marketing campaigns can be carried out, continuously improve and enhance the marketing process, create a traceable marketing closed-loop management, and realize refined operation of scenarios.

[0099] In some embodiments, the user portrait model and marketing strategies can be continuously optimized according to user feedback data and the evaluation results of marketing campaign effectiveness. Specifically, machine learning algorithms are used to automatically adjust model parameters and improve the prediction accuracy of the model. Optimize the screening conditions for target customers, improve the marketing content generation algorithm, and improve marketing copywriting.

[0100] Specifically, the marketing effect analysis support method is as follows:

[0101] (1) Display the effectiveness of marketing campaigns through data visualization tools (such as Chart.js, ReCharts, etc.), including key indicators such as conversion rates and ROI.

[0102] (2) Provide multi-dimensional data analysis reports to help business personnel evaluate the effectiveness of marketing campaigns.

[0103] For the data analysis of existing and new users, analyze customer characteristics through multiple data dimensions such as age, customer assets, transaction preferences, and product holdings, form a user portrait, and convert it into a precision marketing customer model. At the same time, through intelligent components, discover the deviation and preferences of user data to form tag data. After the customer data is labeled, it is output to different marketing strategies according to customer characteristics and data performance.

[0104] Sort out the business and platform indicators required for data-driven operation, meet the requirements of data-driven operation indicators, and complete work such as data aggregation, calculation, page display, and performance optimization. Build a data operation indicator system for analyzing the operation of various platforms and businesses.

[0105] Through the marketing analysis effect, the following optimization solutions are supported:

[0106] (1) Use the data analysis results to help business personnel optimize marketing strategies and improve the accuracy and effectiveness of marketing campaigns.

[0107] (2) Support A / B testing and multivariate testing to find the best marketing plan through experiments.

[0108] Specifically, model optimization includes the following two steps:

[0109] (1) Optimization before going online: feature extraction, sample sampling, and parameter adjustment;

[0110] (2) After the launch, the model is improved based on actual A / B testing and suggestions from business personnel.

[0111] In this embodiment, it is possible to connect and integrate customer data across systems, make full use of the results of customer feature mining and analysis, and achieve in-depth marketing, precision marketing, cross-marketing and real-time marketing. By using big data, machine learning and artificial intelligence technologies, it is possible to automatically track marketing results, continuously optimize marketing strategies, and improve marketing results.

[0112] In this embodiment, user management functions can also be provided, such as user registration, login, and permission management; system monitoring functions can be provided, such as monitoring system operating status, data storage, user access, etc.; system operation logs can be recorded to facilitate user troubleshooting and system maintenance. Specifically, the unified data application platform can provide a unified data analysis application platform, support unified authentication users integrated with the AD domain, support user permissions and data usage permission management, and provide data usage statistics functions. It also has good scalability, can integrate a variety of BI visualization tools, and can integrate multiple data applications. It also integrates management cockpits, customer analysis applications, risk management applications, business product analysis, operational efficiency analysis, application data re-entry and other applications.

[0113] The embodiment of the present invention also provides a method for constructing a user portrait model, including: first obtaining user data of the user, and determining the portrait label of the user data based on a pre-constructed multi-dimensional label system; wherein the label system includes at least: behavior labels, interest labels; then obtaining the user's real-time behavior data, and updating the portrait label based on the real-time behavior data; finally, training the machine learning model based on the user data and the portrait label to obtain a user portrait model.

[0114] In specific implementation, machine learning and data mining algorithms can be adopted to deeply explore the potential information and patterns in customer data, and construct a multi-dimensional label system, including behavior labels (such as purchase frequency, preferred products), interest labels (such as favorite brands, consumption habits), etc., to achieve the refined construction of customer portraits. Meanwhile, a dynamic update mechanism is introduced. Through real-time data collection and analysis technologies, the behavior changes of customers are captured in a timely manner. Using automated scripts or rules, the portrait labels are automatically updated according to the behavior changes of customers (such as purchase records, browsing records, credit score changes, etc.) to ensure the accuracy and timeliness of the portraits.

[0115] In the embodiments of the present invention, when constructing a user portrait, the user data inside and outside the financial institution can be sorted out and integrated, and the user data and behaviors can be analyzed to form a unified view of users, including the unified views of corporate and personal customers, construct a unified customer label system, and construct customer portraits that support application scenarios such as customer insight, channel optimization, precision marketing, and risk prevention. Based on the customer portrait system, business personnel can achieve customer management and customer group management on the visual interface.

[0116] The label data is uniformly managed to achieve a unified view of customer data, metrics, and labels in different systems. Through visual operations, menus, and operation permission management, visual operations of functions such as metric production, label management, population management, and population insight are provided for business, marketing, and operation personnel in different scenarios, helping business personnel quickly understand customer metrics and find target customer groups, and providing data support for work such as customer group tracking and operation.

[0117] In the embodiments of the present invention, a user portrait model can be constructed through a marketing modeling platform. The marketing modeling platform extracts the source system data file from the ODS, and then directly imports the file into the modeling platform environment to complete the data ETL work in the SAS modeling environment to meet the requirements of the modeling theme model. The final modeling result is provided to the downstream system (CRM) application in the form of a unified text format file periodically.

[0118] For the training of the marketing model, according to the extracted features, a sample wide table is formed and input into the classification model. Here, nearly a hundred distributed algorithms provided by the machine learning component Discover of the TDH platform are selected for modeling and training, and the high-order cross characteristics of the features are also used for the prediction and analysis of recommendations.

[0119] Data automatic smoothing segmentation technology, variable IV and WOE calculation system, variable automatic derivation, variable automatic elimination and selection and other technical means are introduced. The relevant models have high accuracy and strong replicability, forming a set of modeling theory systems that can be quickly implemented. Compared with traditional data mining methods such as ACRM customer identification and analysis, event marketing, and precision marketing in the past, the model has high accuracy and stability, low overall input cost, strong model continuous optimization ability, and effectively avoids common problems such as overfitting and underfitting. Among them, the automatic smoothing segmentation technology mainly controls the peak value of the segmentation result, the difference degree of the segmentation, the number of segmentation intervals, etc., and automatically adjusts the monotonicity of the data distribution and the smoothness between segments, solves the problem of inaccurate manual binning accuracy and easy overfitting, improves the accuracy of the model, and also improves the modeling efficiency.

[0120] For ease of understanding, the embodiment of the present invention also provides another credit marketing method based on user portraits. See Figure 2 as shown, including the following steps:

[0121] 1. Data preparation: Collect and integrate existing user data, including user demographic information, behavioral data, transaction data, etc.

[0122] 2. Feature engineering: Preprocess and extract features from user data to construct user feature vectors.

[0123] 3. Model training: Use machine learning algorithms to train a population selection model based on historical marketing data and user feature vectors.

[0124] 4. Population selection: Input the target user feature vector into the trained model to predict the population category to which the user belongs.

[0125] 5. Result evaluation: Evaluate the population selection result and continuously optimize the model according to the evaluation result.

[0126] The above method provided by the embodiments of the present invention integrates customer data scattered in various business systems through a data collection module and a data cleaning module, breaks data silos, and realizes unified management of customer data; conducts model training based on a large amount of user data, driven by data, to improve the accuracy of population selection; depicts user characteristics from multiple dimensions, constructs a more comprehensive and accurate user portrait, and improves the accuracy of population selection. Based on machine learning algorithms, a precise customer portrait is constructed to help financial institutions deeply understand customer needs and achieve precise marketing. By using artificial intelligence technology, the automatic generation and execution of marketing activities are realized, the marketing efficiency is improved, and the operation cost is reduced. By using a risk warning model, customer behaviors are monitored in real time, potential risk customers are identified, and risk control measures are taken in a timely manner to reduce credit risks. By using machine learning algorithms, automatic and intelligent population selection is realized, manual intervention is reduced, and the selection efficiency is improved. Machine learning algorithms with strong interpretability are adopted to facilitate understanding of population division rules and improve the credibility of the model.

[0127] For the credit marketing method based on user portraits provided in the foregoing embodiments, the embodiments of the present invention also provide a credit marketing device based on user portraits. Refer to Figure 3 the structural schematic diagram of a credit marketing device based on user portraits shown in

[0128] A feature extraction module 301, configured to obtain user data of existing users and perform feature extraction on the user data to obtain user feature vectors; wherein, the user data at least includes user basic information, behavior data, credit data, and user interest data;

[0129] A user portrait generation module 302, configured to generate a user portrait of each existing user based on the user feature vectors and a pre-constructed user portrait model;

[0130] A target user screening module 303, configured to screen existing users based on the user portrait and pre-set user screening conditions for the target product to obtain target users;

[0131] A marketing content generation module 304, configured to generate marketing content for target users based on the user portraits of target users and marketing activities of the target product;

[0132] A marketing content pushing module 305, configured to determine a marketing channel based on the user portrait of the target user and push the marketing content to the corresponding target users based on the marketing channel.

[0133] The above-mentioned credit marketing device based on user portraits provided by the present invention can construct user portraits according to multi-dimensional user data, improving the accuracy of user portraits. At the same time, based on the user portraits and user screening conditions for target products, the existing users are screened to obtain target users, and then personalized marketing content for each target user is generated and pushed to the corresponding target users through marketing channels, thereby improving the marketing effect and accuracy of credit products.

[0134] In one implementation, the above-mentioned feature extraction module 301 is specifically configured to extract features from user data to obtain initial user features; screen the initial user features based on a preset feature selection algorithm to obtain user features, and encode the user features to obtain user feature vectors; wherein, the preset features include one of the following: filtering method, wrapping method, and embedding method.

[0135] In one implementation, the above-mentioned user portrait generation module 302 is specifically configured to construct a user portrait hierarchical architecture based on the product requirements of the target product; construct combined feature vectors for each level of user portraits based on the user feature vectors and a pre-constructed user portrait model; calculate the information entropy of the combined feature vectors and the user feature vectors, and determine the user features of each level of user portraits based on the information entropy to generate user portraits for each level.

[0136] In one implementation, the above-mentioned marketing content pushing module 305 is specifically configured to classify the target users based on the basic information of the target users to obtain targeted designated users and excluded users, and push the marketing content to the targeted designated users; determine secondary push users based on the business data of the target users, and push the marketing content to the secondary push users.

[0137] In one implementation, the above-mentioned device further includes an optimization module for obtaining execution information of the marketing activity; wherein, the execution information includes one or more of the following: the sending time of the marketing content, the sent content, and the sending channel; obtaining feedback information of the target users on the marketing activity; wherein, the feedback information includes one or more of the following: click-through rate, conversion rate, and return on investment; optimizing the marketing activity and user portraits based on the execution information and the feedback information.

[0138] In one implementation, the above-mentioned marketing content generation module 304 is specifically configured to construct prompts for a large language model according to the user portraits of the target users and the marketing activities of the target products; input the prompts into the large language model to generate marketing content for the target users.

[0139] In one embodiment, the above device further includes: a model construction module, configured to obtain user data of a user, and determine portrait labels of the user data based on a pre-constructed multi-dimensional label system; wherein the label system at least includes: behavior labels, interest labels; obtain real-time behavior data of the user, and update the portrait labels based on the real-time behavior data; train a machine learning model based on the user data and the portrait labels to obtain a user portrait model.

[0140] It should be noted that for the device provided in the embodiments of the present invention, its implementation principle and the resulting technical effects are the same as those of the foregoing method embodiments. For the sake of brief description, for the parts not mentioned in the device embodiments, reference may be made to the corresponding content in the foregoing method embodiments.

[0141] The embodiments of the present invention also provide an electronic device. Specifically, the electronic device includes a processor and a storage device; a computer program is stored on the storage device, and when the computer program is run by the processor, it executes the method described in any one of the above embodiments.

[0142] Figure 4 FIG. 10 is a schematic structural diagram of an electronic device provided in an embodiment of the present invention. The electronic device 100 includes: a processor 40, a memory 41, a bus 42, and a communication interface 43. The processor 40, the communication interface 43, and the memory 41 are connected through the bus 42; the processor 40 is configured to execute an executable module stored in the memory 41, such as a computer program.

[0143] Among them, the memory 41 may include a high-speed random access memory (RAM, Random Access Memory), and may also include a non-volatile memory, such as at least one disk memory. Through at least one communication interface 43 (which may be wired or wireless), a communication connection is established between the system network element and at least one other network element, and the Internet, a wide area network, a local area network, a metropolitan area network, etc. can be used.

[0144] The bus 42 may be an ISA bus, a PCI bus, an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, Figure 4 only a bidirectional arrow is used in FIG. 10, but it does not mean that there is only one bus or one type of bus.

[0145] Among them, the memory 41 is used to store a program. After receiving an execution instruction, the processor 40 executes the program. The method executed by the device defined by the flow process disclosed in any one of the foregoing embodiments of the present invention can be applied to the processor 40 or implemented by the processor 40.

[0146] The processor 40 may be an integrated circuit chip with the ability to process signals. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in the processor 40 or the instructions in the form of software. The above-mentioned processor 40 may be a general-purpose processor, including a central processing unit (CPU for short), a network processor (NP for short), etc.; it may also be a digital signal processor (DSP for short), an application specific integrated circuit (ASIC for short), a field-programmable gate array (FPGA for short), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present invention can be directly embodied as being executed and completed by a hardware decoding processor, or executed and completed by a combination of hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory 41, and the processor 40 reads the information in the memory 41 and combines its hardware to complete the steps of the above method.

[0147] The computer program product of the readable storage medium provided by the embodiments of the present invention includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the methods described in the foregoing method embodiments. For the specific implementation, reference can be made to the foregoing method embodiments, and details are not described herein again.

[0148] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs.

[0149] Finally, it should be noted that the above-mentioned embodiments are only specific embodiments of the present invention, used to illustrate the technical solutions of the present invention, rather than limiting them. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: any person skilled in the art within the technical scope disclosed by the present invention can still modify the technical solutions recorded in the foregoing embodiments or can easily think of changes, or perform equivalent replacements on some of the technical features; and these modifications, changes, or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A credit marketing method based on user portraits, characterized in that, Including: Obtain the user data of existing users, and extract features from the user data to obtain user feature vectors; wherein, the user data at least includes user basic information, behavior data, credit data, and user interest data; Generate user portraits for each existing user based on the user feature vectors and a pre-constructed user portrait model; Screen the existing users based on the user portraits and pre-set user screening conditions for the target product to obtain target users; Generate marketing content for the target users based on the user portraits of the target users and the marketing activities of the target product; Determine the marketing channels based on the user portraits of the target users, and push the marketing content to the corresponding target users based on the marketing channels.

2. The method according to claim 1, characterized in that, Extracting features from the user data to obtain user feature vectors includes: Extract initial user features from the user data; Screen the initial user features based on a preset feature selection algorithm to obtain user features, and encode the user features to obtain user feature vectors; wherein, the preset features include one of the following: filtering method, wrapper method, and embedding method.

3. The method according to claim 1, wherein Generating user portraits for each existing user based on the user feature vectors and a pre-constructed user portrait model includes: Construct a user portrait hierarchical architecture based on the product requirements of the target product; Construct combined feature vectors for each level of user portraits based on the user feature vectors and a pre-constructed user portrait model; Calculate the information entropy of the combined feature vectors and the user feature vectors, and determine the user features of each level of user portraits based on the information entropy to generate user portraits for each level.

4. The method according to claim 1, wherein Pushing the marketing content to the corresponding target users based on the marketing channels includes: Classify the target users based on the basic information of the target users to obtain targeted designated users and excluded users, and push the marketing content to the targeted designated users; Determine secondary push users based on the business data of the target users, and push the marketing content to the secondary push users.

5. The method according to claim 1, characterized in that, After pushing the marketing content to the corresponding target users based on the marketing channels, it further includes: Obtain the execution information of the marketing activities; wherein, the execution information includes one or more of the following: the sending time, sending content, and sending channel of the marketing content; Obtain the feedback information of the target users on the marketing activities; wherein, the feedback information includes one or more of the following: click-through rate, conversion rate, and return on investment; Optimize the marketing activities and the user portraits based on the execution information and the feedback information.

6. The method according to claim 1, wherein Generating marketing content for the target users based on the user portraits of the target users and the marketing activities of the target product includes: Construct prompt words for the large language model according to the user portraits of the target users and the marketing activities of the target product; Input the prompt words into the large language model to generate marketing content for the target users.

7. The method according to claim 1, wherein The construction of the user portrait model includes: Obtain the user data of the user, and determine the portrait tags of the user data based on a pre-constructed multi-dimensional tag system; wherein, the tag system at least includes: behavior tags, interest tags; Obtain the real-time behavior data of the user, and update the portrait tags based on the real-time behavior data; Train a machine learning model based on the user data and the portrait tags to obtain a user portrait model.

8. A credit marketing device based on user portraits, characterized in that Includes: A feature extraction module, configured to obtain the user data of the existing users, and perform feature extraction on the user data to obtain user feature vectors; wherein, the user data at least includes user basic information, behavior data, credit data, and user interest data; A user portrait generation module, configured to generate a user portrait for each existing user based on the user feature vectors and a pre-constructed user portrait model; A target user screening module, configured to screen the existing users based on the user portrait and the pre-set user screening conditions for the target product to obtain target users; A marketing content generation module, configured to generate marketing content for the target users based on the user portraits of the target users and the marketing activities of the target product; A marketing content push module, configured to determine a marketing channel based on the user portrait of the target user, and push the marketing content to the corresponding target users based on the marketing channel.

9. An electronic device, characterized in that, Includes a processor and a memory, the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it executes the steps of the method according to any one of claims 1 to 7 above.

Images