AI credit marketing management system based on multi-modal data fusion

Abstract

The invention relates to the technical field of credit marketing, in particular to an AI credit marketing management system based on multi-modal data fusion. Multi-source heterogeneous data containing social media, credit transactions and the like are obtained through a web crawler tool, an API interface and the like, the social media and credit transaction data are integrated to construct customer feature vectors, so that the problem of demand misjudgment caused by data islands is solved, real-time compliance constraint data are generated in combination with financial market and credit transaction data, and the real-time compliance constraint data is obtained. The method comprises the following steps: generating a personalized credit marketing strategy based on data to respond to market fluctuation by using'parameter dynamic generation + combinatorial optimization + preference weighting 'based on the data so as to meet diversified requirements of customers, finally calculating marketing accuracy by obtaining a multi-dimensional matching index and interactive feedback data, and replacing subjective judgment by adopting'historical matching + real-time prediction' three-dimensional evaluation so as to improve marketing accuracy. And comparing the marketing accuracy with a preset threshold value, and if the marketing accuracy does not reach the standard, adjusting a marketing strategy so as to realize intelligent upgrading of credit marketing.

Description

Technical Field

[0001] The present invention relates to the technical field of credit marketing, and particularly to an AI credit marketing management system based on multi-modal data fusion. Background Art

[0002] Credit marketing refers to the commercial activities of financial institutions (such as banks, consumer finance companies, etc.) centered around credit products (such as loans, credit cards, etc.). Through a series of strategies such as market research, product design, and channel promotion, financial services are promoted to target customer groups to achieve profit goals and meet the capital needs of customers. Its essence is to accurately match financial resources with customer needs. Against the backdrop of the booming development of fintech, the field of credit marketing is facing profound changes and innovation needs. With the acceleration of digital transformation, customer demands for credit products are becoming increasingly diverse and personalized. The differences between personal consumption scenarios and enterprise operation needs require financial institutions to accurately grasp customer needs.

[0003] Existing credit marketing systems have significant limitations in processing multi-source heterogeneous data. Usually, they only stay at the simple level of feature splicing and fail to deeply explore the potential associations between different modal data. Due to the one-sidedness and lack of depth in data fusion, existing marketing strategies are seriously out of touch with actual customer needs, and the marketing accuracy is extremely low, resulting in poor marketing effects. Summary of the Invention

[0004] The main object of the present invention is to provide an AI credit marketing management system based on multi-modal data fusion, aiming to solve the technical problems in the prior art.

[0005] The present invention proposes an AI credit marketing management system based on multi-modal data fusion, including: An acquisition module for acquiring multi-modal data, where the multi-modal data includes social media data, credit transaction data, financial market data, and customer interaction data, and obtaining a customer feature vector according to the social media data and the credit transaction data; A constraint generation module for obtaining compliance constraint data according to the financial market data and the credit transaction data; A strategy synthesis module for obtaining a credit marketing strategy according to the compliance constraint data, the customer feature vector, and the credit transaction data; An evaluation module for obtaining multi-dimensional matching indicators and interaction feedback data according to the credit marketing strategy and the customer interaction data, and obtaining marketing accuracy according to the multi-dimensional matching indicators and the interaction feedback data; A judgment module for judging whether the marketing accuracy is greater than a preset threshold; If the marketing accuracy is greater than the preset threshold, it is determined that the credit marketing strategy meets customer needs; If the marketing accuracy is not greater than a preset threshold, it is determined that the credit marketing strategy does not meet the customer needs, and then the credit marketing strategy is adjusted according to the multi-dimensional matching indicators until the credit marketing strategy meets the customer needs.

[0006] Preferably, the acquisition module includes: A feature extraction unit, configured to obtain consumer intention text data and consumer intention behavior data according to the social media data, and obtain a plurality of consumer intention targets, consumer intention intensities, and consumer intention timestamps according to the consumer intention text data; A feature analysis unit, configured to obtain a consumption cycle and a risk feature value according to the credit transaction data; A demand acquisition unit, configured to obtain the real-time intention intensity of each consumer intention target according to the consumer intention intensity and the consumer intention timestamp, and obtain a demand feature value of each consumer intention target according to the real-time intention intensity and the consumption cycle; A behavior coding unit, configured to obtain a behavior feature value of each consumer intention target according to the consumer intention behavior data and the consumer intention timestamp; A feature fusion unit, configured to obtain a customer feature vector according to a plurality of the demand feature values, behavior feature values, and risk feature values.

[0007] Preferably, the feature analysis unit in the acquisition module includes: A time series segmentation sub-unit, configured to obtain transaction sequence data and original credit investigation data according to the credit transaction data, and obtain a plurality of transaction windows and transaction category groupings according to the transaction sequence data; A fund analysis sub-unit, configured to obtain corresponding statistical feature data according to each of the transaction windows, and obtain the cash flow health according to the statistical feature data; A cycle acquisition sub-unit, configured to perform a Fourier transform on each of the transaction category groupings to obtain a corresponding consumption cycle; A credit investigation analysis sub-unit, configured to obtain credit investigation feature data according to the original credit investigation data, and obtain a credit investigation risk level according to the credit investigation feature data; A risk quantification sub-unit, configured to obtain a risk feature value according to the credit investigation risk level and the cash flow health.

[0008] Preferably, the constraint generation module includes: A policy analysis unit, configured to obtain policy constraint rules and interest rate fluctuation indicators according to the financial market data, where the policy constraint rules include loan limit rules, interest rate control rules, and repayment time rules, and obtain a regulatory interest rate floor, a static interest rate rule, and a dynamic interest rate rule according to the interest rate control rules; A market perception unit, configured to obtain an interest rate volatility and a volatility threshold according to the interest rate fluctuation indicator, and obtain a market adjustment value according to the interest rate volatility and the volatility threshold; An interest rate limit unit, configured to obtain a lower limit of the loan interest rate according to the interest rate volatility, the volatility threshold, the regulatory lower limit of the interest rate, the static interest rate rule, and the dynamic interest rate rule; A quota limit unit, configured to obtain a debt repayment ability evaluation value according to the credit transaction data, and obtain a basic loan quota and a loan multiple quota according to the debt repayment ability evaluation value and the loan quota rule; A repayment limit unit, configured to obtain a maximum loan amount according to the basic loan quota, the loan multiple quota, and the market adjustment value, and obtain a lower limit of the repayment time according to the repayment time rule and the market adjustment value.

[0009] Preferably, the strategy synthesis module includes: An element generation unit, configured to obtain credit element parameters according to the compliance constraint data and the customer feature vector, where the credit element parameters include multiple loan recommended interest rates, repayment suggestion periods, and loan suggestion amounts; A combination unit, configured to perform permutation and combination on the multiple loan recommended interest rates, repayment suggestion periods, and loan suggestion amounts to obtain multiple marketing strategy combinations; A matching degree acquisition unit, configured to obtain a single-dimensional matching degree according to the credit transaction data, the social media data, and the credit element parameters; A weight analysis unit, configured to obtain preference feedback data according to the customer interaction data, and perform a correlation analysis on the credit elements and the preference feedback data to obtain multiple single-dimensional weights; A preference evaluation unit, configured to obtain a customer preference value of each marketing strategy combination according to the multiple single-dimensional matching degrees and the corresponding single-dimensional weights; A screening unit, configured to screen the multiple marketing strategy combinations according to the customer preference value to obtain a credit marketing strategy.

[0010] Preferably, the matching degree acquisition unit in the strategy synthesis module includes: An interval acquisition subunit, configured to obtain a preference parameter interval according to the credit transaction data; A preference acquisition subunit, configured to obtain a historical preference feature according to the social media data, and obtain a credit parameter preference value according to the historical preference feature and the preference parameter interval, where the credit parameter preference value includes an interest rate preference value, a repayment preference period, and an amount preference value; A preference mapping subunit, configured to obtain corresponding single-dimensional matching degrees according to the credit marketing strategy, interest rate preference value, repayment preference period, and amount preference value, where the single-dimensional matching degrees include multiple loan interest rate matching degrees, repayment period matching degrees, and loan amount matching degrees.

[0011] Preferably, the evaluation module includes: A scenario matching unit, configured to obtain a marketing scenario label according to the credit marketing strategy, and obtain historical matching data according to the marketing scenario label and the customer interaction data, where the historical matching data includes historical strategy parameters and interaction feedback data; A multi-dimensional analysis unit, configured to obtain multi-dimensional matching indicators according to the credit marketing strategy and the historical strategy parameters, where the multi-dimensional matching indicators include parameter adaptability, scenario adaptability, and repayment period adaptability; A strategy evaluation unit, configured to obtain a strategy matching degree according to the parameter adaptability, scenario adaptability, and repayment period adaptability; An interaction prediction unit, configured to obtain the number of interaction behaviors, the number of reached users, the number of conversion behaviors, and the interaction time according to the interaction feedback data, and obtain an expected interaction probability according to the number of interaction behaviors, the number of reached users, and the interaction time; An accuracy acquisition unit, configured to obtain an expected conversion rate according to the number of interaction behaviors, the number of conversion behaviors, and the interaction time, and obtain a marketing accuracy according to the strategy matching degree, the expected conversion rate, and the expected interaction probability.

[0012] Preferably, the judgment module includes: A parameter optimization unit, configured to obtain a matching degree threshold, and obtain credit parameters to be optimized according to the matching degree threshold and the multi-dimensional matching indicators; A gradient analysis unit, configured to obtain an accuracy gradient according to the multi-dimensional matching indicators and the customer preference value; A step size configuration unit, configured to obtain an initial adjustment step size according to the financial market data, and obtain a step size constraint upper limit according to the policy constraint rule; A dynamic adjustment unit, configured to obtain an adaptive step size according to the initial adjustment step size, the step size constraint upper limit, and the accuracy gradient, and obtain a credit parameter adjustment value according to the credit parameters to be optimized and the adaptive step size; An iteration unit, configured to obtain a credit marketing adjustment strategy according to the credit parameter adjustment value and the credit marketing strategy, and return the credit marketing adjustment strategy as the credit marketing strategy to the step of obtaining the marketing accuracy according to the credit marketing strategy and the customer interaction data.

[0013] The present invention also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the memory executes the computer program, it is used to manage the operation of the above-mentioned AI credit marketing management system based on multimodal data fusion.

[0014] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is used to manage the operation of the above-mentioned AI credit marketing management system based on multimodal data fusion.

[0015] The beneficial effects of the present invention are as follows: The present invention uses a web crawler tool to obtain multimodal data according to the API interfaces of social media platforms, credit agencies, etc. By integrating social media data and credit transaction data, a customer feature vector that combines customer needs and risk characteristics is constructed, thus solving the problem of "misjudgment of needs caused by data islands". Then, by combining financial market data and credit transaction data, real-time updated compliance constraint data is generated to respond to market fluctuations in real time. Then, based on the compliance constraint data, customer feature vector and credit transaction data, a personalized credit marketing strategy tailored to customers is generated to solve the contradiction between "single strategy and diverse needs" in traditional solutions. Then, according to the multi-dimensional matching indicators and interaction feedback data obtained from the credit marketing strategy and customer interaction data, the marketing accuracy is further obtained. Through the marketing accuracy, the traditional subjective judgment mode can be replaced, and the marketing effect can be objectively predicted by data means to avoid strategy misjudgment caused by experience deviation. Finally, the marketing accuracy is compared with a preset threshold. If the marketing accuracy is greater than the preset threshold, it is determined that the credit marketing strategy meets the customer's needs; if the marketing accuracy is not greater than the preset threshold, it is determined that the credit marketing strategy does not meet the customer's needs, and the credit marketing strategy is adjusted according to the multi-dimensional matching indicators until the credit marketing strategy meets the customer's needs, which is conducive to the intelligent upgrade of credit marketing. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic structural diagram of a system according to an embodiment of the present invention.

[0017] Figure 2 It is a schematic internal structure diagram of a computer device according to an embodiment of the present application.

[0018] The realization, functional characteristics and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0020] Such as Figure 1As shown, the present application provides an AI credit marketing management system based on multi-modal data fusion. Refer to Figure 1 , in this management system, An acquisition module, configured to acquire multi-modal data, where the multi-modal data includes social media data, credit transaction data, financial market data, and customer interaction data, and acquire a customer feature vector according to the social media data and the credit transaction data; A constraint generation module, configured to acquire compliance constraint data according to the financial market data and the credit transaction data; A strategy synthesis module, configured to acquire a credit marketing strategy according to the compliance constraint data, the customer feature vector, and the credit transaction data; An evaluation module, configured to acquire multi-dimensional matching metrics and interaction feedback data according to the credit marketing strategy and the customer interaction data, and acquire marketing accuracy according to the multi-dimensional matching metrics and the interaction feedback data; A judgment module, configured to judge whether the marketing accuracy is greater than a preset threshold; If the marketing accuracy is greater than the preset threshold, it is determined that the credit marketing strategy meets the customer needs; If the marketing accuracy is not greater than the preset threshold, it is determined that the credit marketing strategy does not meet the customer needs, and then the credit marketing strategy is adjusted according to the multi-dimensional matching metrics until the credit marketing strategy meets the customer needs.

[0021] As described in the functions of the above modules, the present invention uses a web crawler tool to obtain multimodal data according to the API interfaces of social media platforms, credit agencies, etc. Among them, multimodal data refers to a set of information from different sources with different data types and structures, covering various forms such as text, images, audio, and structured data, including social media data, credit transaction data, financial market data, and customer interaction data. Among them, social media data refers to the behavioral data generated by users on social platforms, including unstructured data such as text content, image information, and interaction records. Credit transaction data refers to the historical records of users in financial transactions, which belong to structured data, including credit information, income and expenditure statements, repayment records, etc. Financial market data refers to real-time data related to the macro financial environment, including policies and regulations, interest rate fluctuations, market dynamics, etc. Customer interaction data refers to a large number of historical interaction records between users and credit marketing activities, including outreach behaviors, feedback results, conversion data, etc. By integrating social media data and credit transaction data, a customer feature vector that combines customer needs and risk characteristics is constructed. Among them, the customer feature vector refers to abstracting multimodal data into a numerical feature set and representing the comprehensive attributes of customers in vector form. This solution associates cross-modal data, for example, identifying customers with "frequent browsing of high-end real estate on social media + high income in credit investigation + low debt", and accurately matching high-limit mortgage strategies, thus solving the problem of "misjudgment of demand caused by data silos"; Then, by combining financial market data and credit transaction data, compliance constraint data that is updated in real time is generated. Among them, compliance constraint data refers to the constraint data refined to meet regulatory requirements and market rules, so as to respond to market fluctuations in real time. Then, based on the compliance constraint data, customer feature vectors, and credit transaction data, a personalized credit marketing strategy tailored to customers is generated. Among them, the credit marketing strategy refers to a personalized marketing plan generated based on multimodal data fusion, including core elements such as loan recommended interest rates, repayment periods, and loan amounts. The generation of traditional marketing strategies relies on manual experience (such as "uniformly increasing the interest rate by 10%"), without considering customer personalized preferences (such as a certain customer being sensitive to interest rates but accepting a high limit), resulting in a low matching degree between marketing strategies and customer needs. However, this solution uses "dynamic parameter generation + combined optimization + preference weighting", for example, generating a strategy of "high limit + slightly higher interest rate + flexible period" for customers with high income and showing "urgent need for money" on social media, thus solving the contradiction between "single strategy and diverse needs"; Then, based on the credit marketing strategy and customer interaction data, multi-dimensional matching indicators and interaction feedback data are obtained. Among them, the multi-dimensional matching indicators refer to multi-dimensional quantitative indicators used to evaluate the fit between the credit strategy and customer needs, and the interaction feedback data refers to the direct response data of customers to marketing activities obtained from historical interaction records, including behavioral feedback (clicks, consultations) and result feedback (conversions, rejections). And the marketing accuracy is obtained based on the multi-dimensional matching indicators and the interaction feedback data. Among them, the marketing accuracy refers to a quantitative evaluation value that comprehensively considers the multi-dimensional matching indicators and the interaction feedback data to predict the degree of fit between the marketing strategy and customer needs. Through the marketing accuracy, the traditional subjective judgment mode can be replaced, and the marketing effect can be objectively predicted through data-based means, avoiding strategy misjudgment caused by experience deviation. Moreover, this solution uses a three-dimensional evaluation of "historical matching + real-time prediction" (matching degree + interaction + conversion) to solve the problem that the traditional solution only evaluates the effect based on the final conversion rate and ignores the strategy matching process. Finally, the marketing accuracy is compared with a preset threshold. If the marketing accuracy is greater than the preset threshold, it is determined that the credit marketing strategy meets the customer needs; if the marketing accuracy is not greater than the preset threshold, it is determined that the credit marketing strategy does not meet the customer needs, and then the credit marketing strategy is adjusted according to the multi-dimensional matching indicators until the credit marketing strategy meets the customer needs, which is conducive to realizing the intelligent upgrade of credit marketing.

[0022] In one embodiment, the obtaining module includes: A feature extraction unit, configured to obtain consumer intention text data and consumer intention behavior data according to the social media data, and obtain multiple consumer intention targets, consumer intention intensities, and consumer intention timestamps according to the consumer intention text data; A feature analysis unit, configured to obtain a consumption cycle and a risk feature value according to the credit transaction data; A demand obtaining unit, configured to obtain the real-time intention intensity of each consumer intention target according to the consumer intention intensity and the consumer intention timestamp, and obtain the demand feature value of each consumer intention target according to the real-time intention intensity and the consumption cycle; A behavior encoding unit, configured to obtain the behavior feature value of each consumer intention target according to the consumer intention behavior data and the consumer intention timestamp; A feature fusion unit, configured to obtain a customer feature vector according to multiple demand feature values, behavior feature values, and risk feature values.

[0023] As for the functions of the above-mentioned units, the present invention performs semantic parsing on the consumer intention text data through the BERT-wwm model to obtain consumer intention feature data. Among them, the consumer intention text data refers to the text information that can reflect the user's consumption intention and is extracted from social media (such as forums, comment areas, etc.). The consumer intention feature data refers to the structured feature information refined from the consumer intention text. For example, the obtained consumer intention text data includes content such as "Recently, I want to buy a new mobile phone". For this content, a word segmentation tool (such as jieba) is used to segment the text into words. Then, the preprocessed text is converted into word piece tokens (such as "new", "hand", "phone"), and the word piece tokens are input into the BERT-wwm model. Based on the global semantic representation or specific word vectors output by the model, the consumer target in the text is identified through the classification layer to obtain the consumer intention target. Among them, the consumer intention target refers to the specific object or field that the user's consumption intention points to. For example, when the input text is "Recently, I want to buy a new mobile phone", the model outputs "mobile phone" as the consumer intention target. Then, by combining the keyword weights in the text (such as words like "urgently needed", "must" to enhance the intensity), the consumer intention intensity is obtained. Among them, the consumer intention intensity is a quantitative indicator that measures the strength of the user's consumption intention. And the time-related expressions (such as "next month", "end of the year", "this week") are extracted from the text and converted into a unified time format through a time normalization model (such as Date Parser) to obtain the consumer intention timestamp. Among them, the consumer intention timestamp refers to the specific time point when the consumption intention is expressed. Example: If the text "Recently, I want to buy a new mobile phone" is published on January 1, 2025, then the consumer intention timestamp corresponding to "mobile phone" as the consumer intention target is "2025-01-01"; Obtain the consumption cycle and risk characteristic values according to the credit transaction data. Among them, the consumption cycle refers to the time interval rule of the customer's repeated occurrence of the same type of consumption behavior, and the risk characteristic value is a multi-dimensional index that quantifies the customer's credit risk. Then, through the formula " " Calculate to obtain the real-time intention intensity, where, Represents the real-time intention intensity, Represents the consumer intention intensity, Represents the first attenuation coefficient, and the first attenuation coefficient is determined by fitting historical data, Denote the consumption intention time interval, and the consumption intention time interval is the difference between the consumption intention timestamp and the current time. The real-time intention intensity refers to the current moment intention intensity dynamically adjusted by combining the consumption intention intensity and the timestamp. The prior art ignores the timeliness of consumption intention (for example, the car purchase intention three months ago may have nothing to do with the current demand), and directly using the original intensity will lead to misjudgment. This solution reduces the impact of outdated intentions through a time decay mechanism. For example, the car purchase intention with an intensity of 0.9 three months ago is corrected to 0.36, avoiding the triggering of ineffective marketing by "outdated demands". Through methods such as Fourier transform, the periodic pattern of user consumption (such as the 10th of each month being the consumption peak) has been identified, thereby obtaining the consumption cycle peak. Then, according to the formula " ", the cycle matching degree is calculated, where represents the cycle matching degree, represents the consumption intention timestamp, represents the most recent consumption cycle peak, represents the consumption cycle. Then, according to the formula " ", the demand characteristic value is calculated, where represents the demand characteristic value, represents the weight coefficient, which is determined by the consumption scenario (such as in the car loan scenario = 0.7), represents the real-time intention intensity, represents the cycle matching degree. Among them, the demand characteristic value is an index that quantifies the urgency of the user's demand for a specific consumption target. Through mechanisms such as cycle matching, "timeliness feature correction" is achieved, solving the defect of the prior art that "static features cannot reflect the user's dynamic needs", thereby improving the accuracy of marketing timing. Then, time series modeling is performed on the consumption intention behavior data. Among them, the consumption intention behavior data refers to the behavioral trajectory data related to consumption intention on social media, such as the operation records of the user's likes, collections, forwards, following of merchant accounts, browsing of product pages, etc. And LSTM is used to extract sequence features. After standardizing the consumption intention behavior data (such as browsing duration, consultation times, collection quantity), the recent behavior weight is calculated in combination with the timestamp (such as the behavior weight in the recent 7 days is 1, 8 - 30 days is 0.5, and more than 30 days is 0.2). The behavior characteristic value is generated through weighted summation. Among them, the behavior characteristic value is an index used to quantify the activity degree of the user's consumption behavior; Finally, the demand characteristic value, behavior characteristic value, and risk characteristic value corresponding to each consumption intention target are concatenated to obtain the customer characteristic vector. This solution breaks through the limitation of the prior art of "simple concatenation" and realizes cross-modal association mining of social media and credit transaction data through "semantic understanding + time series analysis + graph association modeling". For example, "social media luxury preference" is associated with "high income in credit investigation" through the graph edge weight, thereby solving the "data island" problem.

[0024] In one embodiment, the feature analysis unit in the acquisition module includes: A time series segmentation subunit, used to obtain transaction sequence data and original credit investigation data according to the credit transaction data, and to obtain multiple transaction windows and transaction category groups according to the transaction sequence data; A fund analysis subunit, used for acquiring corresponding statistical characteristic data according to each of the transaction windows, and acquiring cash flow health according to the statistical characteristic data; A cycle acquisition subunit, used for performing Fourier transform on each of the transaction category groups to obtain a corresponding consumption cycle; A credit analysis subunit, used to obtain credit characteristic data according to the original credit data, and to obtain a credit risk level according to the credit characteristic data; The risk quantification subunit is used to obtain a risk characteristic value according to the credit risk level and the cash flow health.

[0025] As for the functions of each sub-unit mentioned above, the present invention obtains transaction sequence data and original credit data from credit transaction data, wherein the transaction sequence data refers to a set of transaction records arranged in chronological order, each record contains fields such as transaction amount, timestamp, transaction type, etc., and the original credit data refers to the unprocessed credit data directly obtained from the credit reporting agency, and the transaction sequence data is processed by sliding window technology to obtain multiple transaction windows and transaction category groupings, wherein the transaction window refers to a fixed time range defined in the time series data (such as the past 30 days, the past 6 months), which is used to intercept local data for statistical analysis, and the transaction category grouping refers to the transaction data according to the nature or scenario of the transaction. Perform classification aggregation, such as dividing by categories such as "catering", "shopping", and "mortgage", and obtain statistical feature data for each window. Statistical feature data refers to the feature values generated after statistical calculation of the original data, including statistics such as mean, variance, and maximum expenditure. Based on these data, the cash flow health is obtained through a time series convolutional network. Cash flow health refers to the quality of a customer's cash flow status evaluated based on transaction sequence data. The transaction category groups are grouped and Fourier transformed to convert the time domain signal into a frequency domain signal. The inverse of the frequency corresponding to the peak in the spectrum can be used as the consumption cycle. The consumption cycle refers to the time interval pattern of customers repeating the same type of consumption behavior. Next, obtain credit information feature data from the original credit information data. The credit information feature data refers to the structured indicators extracted from the original credit information data, including debt ratio, number of overdue times, provident fund base, etc., and perform standardization processing on these data, and map them to credit risk levels through a rule engine. The credit risk level refers to the result of grading personal credit risks through credit information feature data and a scoring model (such as from AAA to D), and perform standardization processing on the credit risk level and the cash flow health degree, and weighted integration according to the business scenario weights to obtain a risk feature value. The risk feature value refers to a multi-dimensional indicator that quantifies the credit risk of customers.

[0026] In one embodiment, the constraint generation module includes: A policy analysis unit, configured to obtain policy constraint rules and interest rate fluctuation indicators according to the financial market data. The policy constraint rules include loan limit rules, interest rate control rules, and repayment time rules, and obtain the regulatory interest rate floor, static interest rate rules, and dynamic interest rate rules according to the interest rate control rules; A market perception unit, configured to obtain an interest rate volatility and a volatility threshold according to the interest rate fluctuation indicator, and obtain a market adjustment value according to the interest rate volatility and the volatility threshold; An interest rate limit unit, configured to obtain a lower limit of the loan interest rate according to the interest rate volatility, volatility threshold, regulatory interest rate floor, static interest rate rules, and dynamic interest rate rules; A quota limit unit, configured to obtain a debt repayment ability evaluation value according to the credit transaction data, and obtain a basic loan quota and a loan multiple quota according to the debt repayment ability evaluation value and the loan limit rule; A repayment limit unit, configured to obtain a maximum loan amount according to the basic loan quota, loan multiple quota, and market adjustment value, and obtain a lower limit of the repayment time according to the repayment time rule and the market adjustment value.

[0027] As for the functions of the above-mentioned various units, the present invention uses NLP technology to parse financial market data in real time, extract structured rules, and obtain policy constraint rules. Among them, the policy constraint rules refer to the mandatory rules that are updated in real time by financial regulatory agencies or market management parties according to the current economic environment and policy orientation, and are used to regulate the business operations of financial institutions, including loan limit rules, interest rate control rules, and repayment time rules. Among them, the loan limit rule refers to the upper limit standard of the borrower's loan amount set by the regulatory agency or within the financial institution; the interest rate control rule refers to the mandatory regulations on the pricing range and adjustment mechanism of loan interest rates; the repayment time rule refers to the specific regulations on loan repayment methods, terms, overdue penalties, etc. And obtain the regulatory interest rate floor, static interest rate rules, and dynamic interest rate rules according to the interest rate control rules. Among them, the regulatory interest rate floor refers to the legal minimum value of the loan interest rate stipulated by the regulatory agency, and the financial institution's pricing shall not be lower than this value. The static interest rate rule refers to the pricing rule where the interest rate remains fixed during the loan term, usually determined based on the benchmark interest rate or the market interest rate plus points. For example: the lower limit of the loan interest rate = regulatory floor + volatility × sensitivity coefficient. The dynamic interest rate rule refers to the pricing mechanism where the interest rate is adjusted regularly according to market indicators (such as LPR, Treasury bond yield). For example: the lower limit of the loan interest rate = regulatory floor + 50BP + volatility × 10BP. This solution realizes the second-level conversion of "policy release → rule digitization" through NLP, thus solving the problems of the traditional solution relying on manual policy interpretation and delayed compliance response; Collect interest rate volatility indicators based on financial market data. Among them, the interest rate volatility indicator refers to the data reflecting the recent change range and frequency of the financial market interest rate, such as the loan offer rate and the volatility threshold. Among them, the volatility threshold refers to the pre-set critical value of interest rate volatility, which is used to judge whether the market volatility has reached the level that requires adjusting the business strategy, and calculate the yield change sequence based on the loan offer rate, and through the standard deviation and the time factor (such as The annualized processing is used to obtain the interest rate volatility. In this solution, by calculating the real-time interest rate volatility, the compliance constraints can dynamically respond to market changes (such as automatically increasing the interest rate floor when the volatility suddenly rises), thus solving the problem that the traditional solution uses static interest rate constraints (such as a fixed interest rate floor) without considering market volatility risks. Then, using a linear mapping model, the market adjustment value is calculated by combining the interest rate volatility and the volatility threshold. Here, the market adjustment value refers to the proportional value for dynamically adjusting loan-related parameters (such as amount, interest rate). Then, it is judged whether the interest rate volatility is greater than the volatility threshold, and the dynamic adjustment logic is triggered: if the interest rate volatility is not greater than the volatility threshold, the loan interest rate floor is obtained according to the regulatory interest rate floor and the static interest rate rule; if the interest rate volatility is greater than the volatility threshold, the loan interest rate floor is obtained according to the regulatory interest rate floor and the dynamic interest rate rule. Here, the loan interest rate floor refers to the lowest interest rate that a financial institution can set when issuing a loan. The traditional solution lacks a market risk linkage mechanism and adopts a "one-size-fits-all" interest rate rule, which cannot differentiate control during market stability periods and turmoil periods, and may lead to the exposure of risk positions during sharp interest rate fluctuations (such as still using a low interest rate floor when the volatility suddenly rises). This solution realizes an adaptive control of "the higher the risk, the stricter the constraint" through threshold-triggered dynamic adjustment; Then, sequence data such as income flow, consumption records, and debt repayments are extracted from the credit transaction data, and these data are processed through a scoring model (such as logistic regression, machine learning model) to obtain the debt repayment ability evaluation value. Here, the debt repayment ability evaluation value refers to a quantitative index for evaluating the borrower's ability to repay the loan on time based on the borrower's credit transaction data (such as income, debt, repayment records). Then, the loan basic amount is obtained according to the loan limit rule. Here, the loan basic amount refers to the benchmark value of the loan amount initially determined based on the loan limit rule, and the loan multiple is obtained according to the debt repayment ability evaluation value. The loan multiple amount is obtained by multiplying the loan multiple and the loan basic amount. Here, the loan multiple amount refers to the upper limit of the loan amount calculated according to the loan multiple. At the same time, according to the formula " ", the market adjustment amount is calculated, where represents the market adjustment amount, represents the loan basic amount, represents the market adjustment value. Then, the maximum value among the loan basic amount, the loan multiple amount, and the market adjustment amount is used as the maximum loan amount. Here, the maximum loan amount refers to the maximum loan amount that the borrower can finally obtain. Finally, the basic repayment time is obtained according to the repayment time rule, and the lower limit of the repayment time is calculated according to the formula " ", where represents the lower limit of the repayment time, represents the basic repayment time, Represents the market adjustment value. Among them, the lower limit of the repayment time refers to the time limit within which the borrower must complete the repayment. The lower limit of the loan interest rate, the maximum loan amount, and the lower limit of the repayment time constitute the compliance constraint data. Through the above steps, this solution has achieved an upgrade from "static compliance" to "dynamic risk control", solved problems such as slow policy response, ignoring market risks, and rough individual assessments in traditional solutions, and constructed a real-time, intelligent, and flexible compliance constraint generation system.

[0028] In one embodiment, the strategy synthesis module includes: An element generation unit for obtaining credit element parameters according to the compliance constraint data and the customer feature vector. Among them, the credit element parameters include multiple loan recommended interest rates, repayment suggestion periods, and loan suggestion amounts; A combination unit for arranging and combining multiple loan recommended interest rates, repayment suggestion periods, and loan suggestion amounts to obtain multiple marketing strategy combinations; A matching degree acquisition unit for obtaining a single-dimensional matching degree according to the credit transaction data, social media data, and the credit element parameters; A weight analysis unit for obtaining preference feedback data according to the customer interaction data and performing a correlation analysis on the credit elements and the preference feedback data to obtain multiple single-dimensional weights; A preference evaluation unit for obtaining the customer preference value of each marketing strategy combination according to multiple single-dimensional matching degrees and the corresponding single-dimensional weights; A screening unit for screening multiple marketing strategy combinations according to the customer preference value to obtain a credit marketing strategy.

[0029] As for the functions of the above-mentioned units, the present invention combines a customer feature vector including a demand feature value, a behavior feature value, and a risk feature value with compliance constraint data including a lower limit of loan interest rate, a maximum loan amount, and a lower limit of repayment time, so as to obtain credit element parameters. Among them, the credit element parameters refer to the parameter intervals corresponding to the core variables (such as loan interest rate, repayment period, and loan amount) that affect loan decisions and customer experience in credit business, including a loan interest rate interval, a repayment period interval, and a loan amount interval. The specific calculation methods are as follows: Loan interest rate interval = [lower limit of loan interest rate, lower limit of loan interest rate * (1 + risk feature value)]; Repayment period interval = [lower limit of repayment time, lower limit of repayment time * (1 + behavior feature value)]; Loan amount interval = [maximum loan amount * demand feature value, maximum loan amount]; And further, corresponding multiple loan recommended interest rates, repayment recommended periods, and loan recommended amounts can be obtained according to the loan interest rate interval, the repayment period interval, and the loan amount interval respectively. Among them, the loan recommended interest rate refers to the loan interest rate recommended for the customer in a personalized manner, the repayment recommended period refers to the repayment frequency recommended for the customer in a personalized manner, and the loan recommended amount refers to the loan amount recommended for the customer in a personalized manner. The traditional solution only prices based on credit investigation data and does not combine the customer's real-time consumption intention (such as the car purchase demand expressed on social media), resulting in the setting of interest rate and amount being divorced from the actual demand. However, this solution realizes the dual drive of "risk pricing + demand pricing" through cross-modal fusion (compliance constraint + consumption intention), and solves the problem of "disconnection between pricing and demand". For example, customers with high consumption intention and low risk can obtain lower interest rates and higher amounts; Then, the multiple loan recommended interest rates, repayment recommended periods, and loan recommended amounts are arranged and combined to obtain multiple marketing strategy combinations. Among them, the marketing strategy combination refers to the combination of marketing means designed for the customer in a personalized manner. This method of systematic arrangement and combination improves the strategy diversity from "single-digit number" to "hundreds", and solves the problem of low matching degree of the traditional "one-size-fits-all" strategy; Obtain the single - dimension matching degree based on credit transaction data, social media data, and credit factor parameters. Here, the single - dimension matching degree refers to the degree of coincidence between the currently recommended credit factor value and the customer's historical preference characteristics. Then, obtain the preference feedback data based on customer interaction data. Here, the preference feedback data refers to the actual response data of the customer to the recommended credit factors (such as interest rate, amount), and perform a correlation analysis (such as Pearson correlation coefficient) on the credit factors and the preference feedback data to obtain multiple single - dimension weights. Here, the single - dimension weight refers to the importance coefficient assigned to each single - dimension matching degree. Then, based on the loan interest rate matching degree, repayment period matching degree, loan amount matching degree, and the corresponding single - dimension weights, obtain the customer preference value of each marketing strategy combination. Here, the customer preference value refers to the numerical value that quantifies the customer's preference degree for specific credit factors. Finally, take the marketing strategy combination corresponding to the largest customer preference value as the credit marketing strategy. This method of sorting by preference value and controlling diversity can achieve the "global optimal" strategy screening, upgrade the strategy customization from "experience - driven" to "data - driven", and is beneficial to solving the industry pain point of low credit strategy matching degree.

[0030] In one embodiment, the matching degree acquisition unit in the strategy synthesis module includes: An interval acquisition subunit, configured to obtain a preference parameter interval according to the credit transaction data; A preference acquisition subunit, configured to obtain historical preference characteristics according to the social media data, and obtain credit parameter preference values according to the historical preference characteristics and the preference parameter interval. Here, the credit parameter preference values include interest rate preference values, repayment preference periods, and amount preference values; A preference mapping subunit, configured to obtain the corresponding single - dimension matching degrees according to the credit marketing strategy, interest rate preference values, repayment preference periods, and amount preference values. Here, the single - dimension matching degrees include multiple loan interest rate matching degrees, repayment period matching degrees, and loan amount matching degrees.

[0031] As the functions of the above - mentioned units, the present invention extracts the customer's preference parameter interval from the credit transaction data. Here, the preference parameter interval refers to the preference range shown by the customer for a certain type of credit parameter (such as interest rate, amount, repayment period) in historical credit behaviors, usually represented in the form of an interval. For example: interest rate preference (such as choosing an interest rate of 5.0% - 5.5% for the past 3 loans), amount habit (the applied amount is between 100,000 and 150,000), and analyzes the text information in the social media data through the BERT - wwm model, such as semantics like "the interest rate is too high" and "it would be great if the amount could reach 200,000", so as to generate historical preference characteristics. Here, the historical preference characteristics refer to the stable preference data shown by the customer in past credit behaviors or consultation behaviors, and are used to depict the customer's demand pattern; Combine historical preference features and preference parameter intervals to obtain credit parameter preference values. Among them, the credit parameter preference value refers to the specific credit parameters preferred by the customer, including interest rate preference value, repayment preference period, and amount preference value. Then, according to the interest rate preference value, repayment preference period, amount preference value, and credit element parameters, obtain the single-dimensional matching degree. Among them, the single-dimensional matching degree refers to the degree of coincidence between the currently recommended credit element value and the customer's historical preference features, including loan interest rate matching degree, repayment period matching degree, and loan amount matching degree. The calculation formula is: ; Among them, represents the loan interest rate matching degree, represents the recommended loan interest rate, represents the interest rate preference value of the loan; If the recommended repayment period is the same as the repayment preference period, the repayment period matching degree is 1. If the recommended repayment period is a multiple of the repayment preference period, the repayment period matching degree is 0.7. If there is any other relationship between the recommended repayment period and the repayment preference period, the repayment period matching degree is 0.3; , among them, represents the loan amount matching degree, represents the recommended loan amount, represents the amount preference value of the loan.

[0032] In one embodiment, the evaluation module includes: A scenario matching unit for obtaining marketing scenario tags according to the credit marketing strategy, and obtaining historical matching data according to the marketing scenario tags and the customer interaction data. Among them, the historical matching data includes historical strategy parameters and interaction feedback data; A multi-dimensional analysis unit for obtaining multi-dimensional matching indicators according to the credit marketing strategy and the historical strategy parameters. Among them, the multi-dimensional matching indicators include parameter adaptation degree, scenario adaptation degree, and repayment period adaptation degree; A strategy evaluation unit for obtaining a strategy matching degree according to the parameter adaptation degree, scenario adaptation degree, and repayment period adaptation degree; An interaction prediction unit for obtaining the number of interaction behaviors, the number of users reached, the number of conversion behaviors, and the interaction time according to the interaction feedback data, and obtaining the expected interaction probability according to the number of interaction behaviors, the number of users reached, and the interaction time; An accuracy acquisition unit for obtaining the expected conversion rate according to the number of interaction behaviors, the number of conversion behaviors, and the interaction time, and obtaining the marketing accuracy according to the strategy matching degree, the expected conversion rate, and the expected interaction probability.

[0033] As for the functions of the above-mentioned units, the present invention maps credit marketing strategies to marketing scenario tags. Among them, the marketing scenario tag refers to a classification tag used to identify the specific scenario characteristics targeted by the credit marketing strategy plan. For example, in the scenario of "small consumer loan - regular interest rate", based on the marketing scenario tag, similar strategy data is extracted from customer interaction data to obtain historical matching data. The historical matching data refers to the corresponding relationship data of "strategy parameters" and "user feedback" stored in historical credit marketing activities, including historical strategy parameters and interaction feedback data. The historical strategy parameters refer to the specific strategy parameters actually adopted in historical credit marketing activities under similar marketing scenario tags, and through the average loan interest rate of historical strategy parameters, historical scenario tags, and average loan amount, the interaction feedback data refers to the behavioral data generated by users during the interaction with credit marketing activities under similar marketing scenario tags, including operation records such as clicks, consultations, application submissions, and rejections. This solution establishes a historical association between strategies and effects through scenario tags, thereby predicting the promotion effect of credit marketing strategies in combination with historical data, and thus solving the problem of repeated pushing of ineffective strategies in traditional methods; Then, according to the credit marketing strategy and historical strategy parameters, multi-dimensional matching indicators are obtained, including sub-indicators such as parameter adaptability, scenario adaptability, and repayment period adaptability, which are used to measure the degree of fit between the current strategy and historical successful strategies. The specific acquisition method is as follows: according to the credit marketing strategy, customized loan interest rates, customized repayment periods, and customized loan amounts for customers are obtained. Then, the parameter adaptability is obtained by calculating the ratio of the average loan interest rate to the customized loan interest rate, and based on the semantic matching of scenario tags (such as the semantic distance between "car loan" and "car installment"), the scenario adaptability is calculated using the Word2Vec model. The repayment period adaptability is obtained by calculating the ratio of the average repayment period to the customized repayment period. Among them, the parameter adaptability refers to the degree of matching between the current credit strategy parameters and historical average parameters, the scenario adaptability refers to the degree of similarity between the current marketing scenario and historical successful scenarios, and the repayment period adaptability refers to the degree of fit between the current repayment period and the user's historical preference period. Then, the strategy matching degree is obtained by weighted summation according to the parameter adaptability, scenario adaptability, and repayment period adaptability. The strategy matching degree refers to the overall degree of fit between the current strategy and historical similar strategies; Then, based on the interactive feedback data, obtain the number of interactive behaviors, the number of reached users, the number of conversion behaviors, and the interaction time. Among them, the number of interactive behaviors refers to the number of effective interactive operations generated by users in the credit marketing activity, such as the statistics of the number of times of clicking on the marketing link, consulting the customer service, viewing the loan details, etc. The number of reached users refers to the number of target users successfully reached by the credit marketing activity, such as the number of users actually delivered through channels such as text messages and APP push. The number of conversion behaviors refers to the number of users who complete the target conversion actions in the marketing activity, such as the number of times of core conversion behaviors such as submitting a loan application and signing a contract. The interaction time refers to the time dimension data of the interaction between the user and the credit marketing content, and calculate the expected interaction probability according to the formula " ". Among them, represents the expected interaction probability, represents the number of interactive behaviors, represents the number of reached users, represents the second attenuation coefficient, represents the time interval, and the time interval refers to the interval between the interaction time and the current time. The expected interaction probability refers to the possibility of predicting that the user will generate an interactive behavior under the current strategy. Calculate the expected conversion rate according to the formula " ". Among them, represents the expected conversion rate, represents the number of conversion behaviors, represents the number of interactive behaviors, represents the second attenuation coefficient, represents the time interval. Among them, the expected conversion rate refers to the probability of predicting that the user will complete the core conversion (such as applying for a loan) under the current strategy. Finally, through the logistic regression model based on customer interaction data, obtain the marketing accuracy according to the strategy matching degree, the expected conversion rate, and the expected interaction probability. The traditional method relies on a single indicator (such as AUC) to evaluate the model and cannot distinguish between "strategy matching problems" and "execution effect problems". For example, for a certain model with an AUC of 0.75, but the strategy matching degree is only 0.5, it is actually a strategy design problem rather than a model defect. And this solution directly locates the root cause of the problem by disassembling the indicators.

[0034] In one embodiment, the judgment module includes: A parameter optimization unit, configured to obtain a matching degree threshold and obtain the credit parameters to be optimized according to the matching degree threshold and the multi-dimensional matching indicators; A gradient analysis unit, configured to obtain the accuracy gradient according to the multi-dimensional matching indicators and the customer preference value; A step size configuration unit, configured to obtain an initial adjustment step size according to the financial market data and obtain a step size constraint upper limit according to the policy constraint rules; A dynamic adjustment unit, configured to obtain an adaptive step size according to the initial adjustment step size, the upper limit of step size constraint, and the precision gradient, and obtain a credit parameter adjustment value according to the credit parameter to be optimized and the adaptive step size; An iteration unit, configured to obtain a credit marketing adjustment strategy according to the credit parameter adjustment value and the credit marketing strategy, and return the credit marketing adjustment strategy as the credit marketing strategy to the step of obtaining the marketing precision according to the credit marketing strategy and the customer interaction data.

[0035] As the functions of the above units, the present invention compares the parameter matching degree, the scenario matching degree, and the repayment period matching degree with the matching degree threshold respectively, and uses the credit parameters with values less than the matching degree threshold as the credit parameters to be optimized. The matching degree threshold is a critical value preset for judging the matching degree of credit parameters with elements such as customer needs and market environment. The credit parameters to be optimized refer to the credit-related parameters that are determined to need adjustment to improve marketing precision. This method of screening through thresholds can not only optimize the parameters with insufficient matching degree but also solve the problem of "unclear optimization direction" in traditional methods; Then, based on the relationship between the multi-dimensional matching index and the customer preference value, a precision function f(x) is constructed, where x is a credit parameter vector (such as loan interest rate, loan amount, repayment period), and the gradient vector of each credit parameter vector is calculated respectively to obtain the precision gradient. The precision gradient is used to measure the sensitivity of adjusting credit parameters to improve precision. According to the gradient vector, the gradient optimization direction can be obtained. For example: the positive gradient direction (such as >0): increasing the interest rate can improve precision; the negative gradient direction (such as <0): decreasing the interest rate can improve precision; Immediately afterwards, the initial adjustment step size is obtained according to the financial market data. The initial adjustment step size is the basic amplitude set initially for each adjustment, and the upper limit of step size constraint is obtained according to the policy constraint rule. The upper limit of step size constraint is the maximum allowable value of the adjustment step size set according to the policy constraint rule (such as regulatory requirements). Through dynamic step size constraint, it is ensured that the optimization conforms to market rules and meets regulatory requirements. For example, the system can automatically identify the current market interest rate trend and avoid excessive downward adjustment of interest rates during the interest rate upward period. Then, according to the formula " " the adaptive step size is calculated, where, represents the adaptive step size, represents the initial adjustment step size, Denote the gradient vector. Among them, the adaptive step size refers to the optimal adjustment amplitude calculated dynamically, and then the credit parameter adjustment value is obtained by combining the gradient optimization direction, the credit parameter to be optimized, and the adaptive step size. The credit parameter adjustment value refers to the value obtained by adjusting the credit parameter to be optimized according to the adaptive step size. For example, if the interest rate gradient is -0.05 (a negative gradient indicates that the interest rate needs to be decreased), the adaptive step size is 0.08%, and the customized loan interest rate is 5.35%, then the interest rate adjustment value is 5.35% - 0.08% × 0.05 = 5.346%. This method of adjustment through the adaptive step size can adopt a large step size for fast iteration when the gradient is large, and a small step size for fine adjustment when the gradient is small; Finally, use the credit parameter adjustment value to replace the credit parameter to be optimized in the credit marketing strategy to obtain the credit marketing adjustment strategy. The credit marketing adjustment strategy refers to the new credit marketing strategy formed by adjusting on the basis of the original credit marketing strategy, and return the credit marketing adjustment strategy back to the step of "obtaining the marketing accuracy according to the credit marketing strategy and customer interaction data" as the credit marketing strategy. This is conducive to achieving the continuous evolution of the strategy through cyclic iteration.

[0036] The present invention also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the memory executes the computer program, it is used to manage the operation of the above-mentioned AI credit marketing management system based on multi-modal data fusion.

[0037] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is used to manage the operation of the above-mentioned AI credit marketing management system based on multi-modal data fusion.

[0038] Those of ordinary skill in the art can understand that all or part of the processes in the above-described embodiment systems can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned modules. Among them, any reference to a memory, storage, database, or other medium provided in this application and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0039] It should be noted that in this text, the term "including", "comprising", or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that includes a series of elements includes not only those elements but also other elements not expressly listed, or elements that are inherent to such process, apparatus, article, or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, apparatus, article, or method that includes the element.

[0040] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent structural or equivalent process transformation made by using the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. An AI credit marketing management system based on multimodal data fusion, characterized in that Including: An acquisition module, configured to acquire multimodal data, where the multimodal data includes social media data, credit transaction data, financial market data, and customer interaction data, and acquire a customer feature vector according to the social media data and the credit transaction data; A constraint generation module, configured to acquire compliance constraint data according to the financial market data and the credit transaction data; A strategy synthesis module, configured to acquire a credit marketing strategy according to the compliance constraint data, the customer feature vector, and the credit transaction data; An evaluation module, configured to acquire multi-dimensional matching metrics and interaction feedback data according to the credit marketing strategy and the customer interaction data, and acquire marketing accuracy according to the multi-dimensional matching metrics and the interaction feedback data; A judgment module, configured to judge whether the marketing accuracy is greater than a preset threshold; If the marketing accuracy is greater than the preset threshold, it is determined that the credit marketing strategy meets the customer needs; If the marketing accuracy is not greater than the preset threshold, it is determined that the credit marketing strategy does not meet the customer needs, and then the credit marketing strategy is adjusted according to the multi-dimensional matching metrics until the credit marketing strategy meets the customer needs.

2. The AI credit marketing management system based on multi-modal data fusion according to claim 1, wherein The acquisition module includes: A feature extraction unit, configured to acquire consumption intention text data and consumption intention behavior data according to the social media data, and acquire multiple consumption intention targets, consumption intention intensity, and consumption intention timestamps according to the consumption intention text data; A feature analysis unit, configured to acquire a consumption cycle and a risk feature value according to the credit transaction data; A demand acquisition unit, configured to acquire the real-time intention intensity of each consumption intention target according to the consumption intention intensity and the consumption intention timestamp, and acquire a demand feature value of each consumption intention target according to the real-time intention intensity and the consumption cycle; A behavior coding unit, configured to acquire a behavior feature value of each consumption intention target according to the consumption intention behavior data and the consumption intention timestamp; A feature fusion unit, configured to acquire a customer feature vector according to multiple demand feature values, behavior feature values, and risk feature values.

3. The AI credit marketing management system based on multi-modal data fusion according to claim 2, characterized in that, The feature analysis unit in the acquisition module includes: A time series segmentation sub-unit, configured to acquire transaction sequence data and original credit investigation data according to the credit transaction data, and acquire multiple transaction windows and transaction category groupings according to the transaction sequence data; A fund analysis sub-unit, configured to acquire corresponding statistical feature data according to each transaction window, and acquire the cash flow health according to the statistical feature data; A cycle acquisition sub-unit, configured to perform Fourier transform on each transaction category grouping to obtain a corresponding consumption cycle; A credit investigation parsing sub-unit, configured to acquire credit investigation feature data according to the original credit investigation data, and acquire a credit investigation risk level according to the credit investigation feature data; A risk quantification sub-unit, configured to acquire a risk feature value according to the credit investigation risk level and the cash flow health.

4. The AI credit marketing management system based on multimodal data fusion according to claim 1, characterized in that The constraint generation module includes: A policy analysis unit for obtaining policy constraint rules and interest rate volatility indicators based on the financial market data, wherein the policy constraint rules include loan limit rules, interest rate control rules, and repayment time rules, and obtaining a regulatory interest rate floor, a static interest rate rule, and a dynamic interest rate rule according to the interest rate control rules; A market perception unit for obtaining an interest rate volatility rate and a volatility threshold based on the interest rate volatility indicator, and obtaining a market adjustment value based on the interest rate volatility rate and the volatility threshold; An interest rate limit unit for obtaining a lower limit of the loan interest rate based on the interest rate volatility rate, the volatility threshold, the regulatory interest rate floor, the static interest rate rule, and the dynamic interest rate rule; A quota limit unit for obtaining a debt repayment ability evaluation value based on the credit transaction data, and obtaining a basic loan quota and a loan multiple quota based on the debt repayment ability evaluation value and the loan limit rules; A repayment limit unit for obtaining a maximum loan amount based on the basic loan quota, the loan multiple quota, and the market adjustment value, and obtaining a lower limit of the repayment time based on the repayment time rule and the market adjustment value; 5. The AI credit marketing management system based on multi-modal data fusion according to claim 4, characterized in that, The strategy synthesis module includes: An element generation unit for obtaining credit element parameters based on the compliance constraint data and the customer feature vector, wherein the credit element parameters include a plurality of loan recommended interest rates, a repayment recommended period, and a loan recommended amount; A combination unit for arranging and combining the plurality of loan recommended interest rates, the repayment recommended periods, and the loan recommended amounts to obtain a plurality of marketing strategy combinations; A matching degree obtaining unit for obtaining a single-dimensional matching degree based on the credit transaction data, the social media data, and the credit element parameters; A weight analysis unit for obtaining preference feedback data based on the customer interaction data, and performing a correlation analysis on the credit elements and the preference feedback data to obtain a plurality of single-dimensional weights; A preference evaluation unit for obtaining a customer preference value of each of the marketing strategy combinations based on the plurality of single-dimensional matching degrees and the corresponding single-dimensional weights; A screening unit for screening the plurality of marketing strategy combinations based on the customer preference value to obtain a credit marketing strategy; 6. The AI credit marketing management system based on multimodal data fusion according to claim 5, characterized in that, The matching degree obtaining unit in the strategy synthesis module includes: An interval obtaining subunit for obtaining a preference parameter interval based on the credit transaction data; A preference obtaining subunit for obtaining a historical preference feature based on the social media data, and obtaining a credit parameter preference value based on the historical preference feature and the preference parameter interval, wherein the credit parameter preference value includes an interest rate preference value, a repayment preference period, and an amount preference value; A preference mapping subunit for obtaining a corresponding single-dimensional matching degree based on the credit marketing strategy, the interest rate preference value, the repayment preference period, and the amount preference value, wherein the single-dimensional matching degree includes a plurality of loan interest rate matching degrees, repayment period matching degrees, and loan amount matching degrees; 7. The AI credit marketing management system based on multi-modal data fusion according to claim 1, characterized in that, The evaluation module includes: A scene matching unit, configured to obtain marketing scene tags according to the credit marketing strategy, and obtain historical matching data according to the marketing scene tags and the customer interaction data, wherein the historical matching data includes historical strategy parameters and interaction feedback data; A multi-dimensional analysis unit, configured to obtain multi-dimensional matching metrics according to the credit marketing strategy and the historical strategy parameters, wherein the multi-dimensional matching metrics include parameter adaptability, scene adaptability, and repayment period adaptability; A strategy evaluation unit, configured to obtain a strategy matching degree according to the parameter adaptability, scene adaptability, and repayment period adaptability; An interaction prediction unit, configured to obtain the number of interaction behaviors, the number of users reached, the number of conversion behaviors, and the interaction time according to the interaction feedback data, and obtain an expected interaction probability according to the number of interaction behaviors, the number of users reached, and the interaction time; An accuracy acquisition unit, configured to obtain an expected conversion rate according to the number of interaction behaviors, the number of conversion behaviors, and the interaction time, and obtain marketing accuracy according to the strategy matching degree, the expected conversion rate, and the expected interaction probability.

8. The AI credit marketing management system based on multimodal data fusion according to claim 5, characterized in that The judgment module includes: A parameter optimization unit, configured to obtain a matching degree threshold, and obtain credit parameters to be optimized according to the matching degree threshold and the multi-dimensional matching metrics; A gradient analysis unit, configured to obtain an accuracy gradient according to the multi-dimensional matching metrics and the customer preference value; A step size configuration unit, configured to obtain an initial adjustment step size according to the financial market data, and obtain a step size constraint upper limit according to the policy constraint rules; A dynamic adjustment unit, configured to obtain an adaptive step size according to the initial adjustment step size, the step size constraint upper limit, and the accuracy gradient, and obtain a credit parameter adjustment value according to the credit parameters to be optimized and the adaptive step size; An iteration unit, configured to obtain a credit marketing adjustment strategy according to the credit parameter adjustment value and the credit marketing strategy, and return the credit marketing adjustment strategy as the credit marketing strategy to the step of obtaining marketing accuracy according to the credit marketing strategy and the customer interaction data.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the memory executes the computer program, it is used to manage the operation of the AI credit marketing management system based on multi-modal data fusion according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it is used to manage the operation of the AI credit marketing management system based on multi-modal data fusion according to any one of claims 1 to 8.

Images