A business adaptability intelligent analysis method and system combining customer portraits

Abstract

The present invention relates to the field of business adaptation technology, specifically to a business adaptation intelligent analysis method and system combined with customer portraits; first, customer data is collected from multiple data sources, then the collected data is processed, customer interest tags are extracted, a customer-interest tag matrix is ​​constructed, and the customer's purchasing power and loyalty are comprehensively evaluated. Then, the customer's portrait is updated according to the interest preferences of similar customers, and then the customer's business demand data is collected, a business demand model is constructed, the customer's demand intensity is quantified, and a demand scoring vector is generated, and the customer portrait is matched with the demand scoring vector, and the adaptation score is calculated to generate personalized recommendations for the customer, and finally the product or service with the highest matching degree with the customer's demand is recommended first, and relevant recommendations are pushed through multiple channels to improve customer satisfaction and conversion rate. The present invention improves resource utilization efficiency and customer satisfaction, thereby achieving more efficient enterprise operations and customer service.

Description

Technical Field

[0001] The present invention relates to the field of business adaptation technology, and in particular to a method and system for intelligent analysis of business adaptability in combination with customer portraits. Background Art

[0002] With the rapid development of big data and artificial intelligence technologies, enterprises have increasing demands for customer management and marketing. Traditional customer management methods usually rely on simple customer stratification and basic marketing strategies, which are often unable to accurately identify subtle differences in customer needs and therefore unable to provide customers with personalized services. At the same time, with the increasing diversification of customer information sources, enterprises obtain a wide variety of customer data, such as basic customer information, behavioral data, social interactions, purchase history, etc. These data contain rich potential value. How to effectively extract meaningful information from these complex data and combine artificial intelligence technology for precise analysis has become the key to improving the competitiveness of enterprises.

[0003] A Chinese invention patent with publication number CN117391788A discloses a business recommendation method, including: obtaining application information, and extracting customer features and product features from the application information, wherein the application information includes structured type information and unstructured type information, and the unstructured information includes text information and image information; determining customer evaluation information according to customer features, determining product evaluation information according to product features, and determining target business according to customer evaluation information and product evaluation information in a preset business warehouse, and generating recommendation information corresponding to the target business; evaluating from both customer and product perspectives at the same time, building a panoramic view of customers and products from different aspects, being able to accurately match customer needs and realize personalized recommendations; and considering structured data and unstructured data, being able to discover richer potential features and improve the accuracy of business recommendations;

[0004] As a data-based customer analysis method, customer portrait technology has been widely used in recent years. Through multi-dimensional data integration and analysis, it can help companies build dynamic customer models, identify customer interests, needs and purchasing potential, and thus achieve precision marketing and personalized recommendations. Existing customer portrait applications are mainly focused on marketing. There is still a lack of effective intelligent analysis methods for how to evaluate the compatibility between customers and businesses and how to optimize businesses based on customers' dynamic needs. Especially in a rapidly changing market environment, how to accurately match customer portraits with specific business needs and guide companies to optimize products, services and marketing strategies to better meet customers' personalized needs has become an urgent problem to be solved. Summary of the invention

[0005] The purpose of the present invention is to address the problems existing in the background technology and to propose a method and system for intelligent analysis of business adaptability combined with customer portraits.

[0006] The technical solution of the present invention is a method for intelligent analysis of business suitability combined with customer portraits, including the following specific implementation steps:

[0007] S1. Collect customer data and generate identification tags for the customer data;

[0008] S2. Verify the accuracy of customer data, and generate a unique customer ID based on the customer data. Use the TF-IDF model to extract interest tags and calculate weights, build a customer-interest tag matrix to generate a customer portrait, and obtain purchase behavior and loyalty scores from customer data, calculate the similarity between customers, generate the final customer portrait, and build the association relationship <unique customer ID, customer portrait>;

[0009] S3. Collect customer demand data, generate a unique customer ID, and build a multi-dimensional business demand model. Use feature engineering to convert demand into quantitative indicators. Use a weighted scoring model to quantify customer business needs. Finally, output a demand scoring vector and build an association relationship <unique customer ID, demand scoring vector>.

[0010] S4. According to the association relationship <unique customer identifier, customer profile> and the association relationship <unique customer identifier, demand score vector>, the customer's profile vector and the demand score vector are matched to evaluate the customer's suitability for specific business requirements, and a suitability score is obtained accordingly;

[0011] S5. Based on the fitness score, generate personalized recommendations through the intelligent recommendation algorithm, obtain the recommendation priority, generate a recommendation list, and make personalized recommendations accordingly;

[0012] S6. Build an automated data processing and model update process, and use batch processing systems and stream data processing frameworks to process and update customer behavior data, business demand data, and customer feedback data in real time, thereby updating user portraits, fitness scores, and intelligent recommendation algorithms.

[0013] Preferably, the process of generating the final customer portrait is as follows:

[0014] S21. Use the TF-IDF model to perform text analysis, automatically extract keywords, calculate the weights of different keywords in the text data, and use keywords with high weights as interest tags;

[0015] S22. Construct a customer-interest tag matrix and generate an interest preference vector using the TF-IDF weighting method:

[0016] ;

[0017] ;

[0018] In the formula, I i represents the interest preference vector of customer i, including the customer's preference for various interest tags; T j represents the interest tag j, i.e. the possible interest point of the customer; w ij represents the interest weight of customer i for interest tag j; tf ij represents the frequency of customer i's behavior on interest tag j; N represents the total number of customers; df j represents the number of customers containing interest tag j, that is, the number of customers who show interest in interest tag j;

[0019] S23. Capture the customer’s most recent purchase time, purchase frequency, and consumption amount from the customer data, build an RFM model, and generate a purchasing power vector P i =[R i , F i , M i ];

[0020] Among them, P i represents the purchasing power vector of customer i; R i represents the most recent purchase time of customer i, that is, the number of days between the most recent purchase and the current time; F i represents the purchase frequency of customer i in a given time period; M i represents the consumption amount of customer i in a given time period;

[0021] S24. Capture the customer's repurchase frequency, total purchase frequency, customer satisfaction, and customer recommendation frequency from the customer data and calculate the loyalty score L i :

[0022] ;

[0023] Where, L i Re represents the loyalty score of customer i; i It represents the proportion of multiple purchases by customer i in a given time period, that is, the number of repeat purchases / total number of purchases; E i represents the evaluation score, i.e., the average satisfaction level of customer i; Ni represents the recommendation behavior score, i.e., the number of times a customer recommends; α, β, γ represent weight coefficients, i.e., the impact of repurchase rate, evaluation score, and recommendation behavior on loyalty;

[0024] S25. Use cosine similarity to calculate the similarity between customers and find the interest preferences of other customers similar to customer i. If the similarity between customer i and customer j is high, the label weight of customer j can be used to update the interest preference vector of customer i to obtain the adjusted interest preference vector of the target customer I'.i :

[0025] ;

[0026] ;

[0027] In the formula, represents the updated interest preference vector of customer i; λ represents the update coefficient, which is used to control the degree of interest preference update; S(i) represents the set of customers similar to customer i; sim(i,j) represents the similarity between customer i and customer j; R ik and R jk denotes the ratings of customer i and customer j on interest tag k respectively; n denotes the total number of interest tags;

[0028] S26, output image vector C i =[B i , I' i , P i , L i ], that is, the customer profile C of the final customer i is obtained i =[B i , I' i , P i , L i ];

[0029] Among them, B i Represents the basic attributes of customer i.

[0030] Preferably, the construction process of the association relationship <unique customer identifier, demand scoring vector> is as follows:

[0031] S31. Collect business demand data;

[0032] S32. Based on the collected business demand data, extract the customer's ID number and generate a unique customer identification ID;

[0033] S33. Build a multi-dimensional business demand model and convert demand into quantifiable indicators through feature engineering: divide customer demand into different categories, quantify each category of demand according to the intensity of customer behavior, and use a weighted scoring model to quantify customer business needs;

[0034] The demand score vector for each dimension is as follows:

[0035] ;

[0036] Where D i represents the business demand score vector of any customer i, that is, the demand intensity of the user in different business demand dimensions; w jrepresents the weight of business requirement j, which reflects the relative importance of different types of requirements; D ij represents the demand score of customer i under demand dimension j;

[0037] S34, build association relationship < unique customer ID i , the business demand score vector D of any customer i i >

[0038] Among them, ID i Represents the unique customer identifier of any customer i.

[0039] Preferably, the generation process of the fitness score is as follows:

[0040] S41. According to the association relationship <unique customer ID, customer portrait> and the association relationship <unique customer ID, demand scoring vector>, the portrait vector C of any customer i is i and the demand score vector D i Make a match;

[0041] S42. Based on the customer's portrait vector and demand score vector, calculate the similarity between the two:

[0042] ;

[0043] In the formula, C i represents the portrait vector of customer i; D i represents the business requirement score vector of customer i; ||C i || represents the modulus of the portrait vector of customer i; ||D i || represents the modulus of the business requirement score vector of customer i;

[0044] S43. Generate a fitness score A by combining the similarity between the customer portrait vector and the demand score vector i :

[0045] ;

[0046] ;

[0047] In the formula, A i represents the suitability score of customer i, that is, the degree of match between customer i and specific business requirements; and represents the adjustment coefficient, which controls the contribution of similarity calculation and other factors to the fitness score; represents an additional function used to enhance the fitness score; represents the historical purchase frequency of a certain type of product by customer i; represents the weighting coefficient, which is used to adjust the impact of historical behavior on the fitness score; represents the number of interactive behaviors of customer i in a given time; represents the weighting coefficient, which is used to control the contribution of activity to the fitness score; Represents the weighting coefficient, which is used to adjust the intensity of time decay; It means that as time goes by, the customer's suitability for a certain demand gradually decreases; represents the attenuation factor, which controls the influence of time distance on the fitness score; t i Indicates the number of days since customer i’s last interaction with a certain type of business demand.

[0048] Preferably, the process of generating the recommendation list is as follows:

[0049] S51. Combine the customer's suitability score with other business indicators to derive the recommendation priority:

[0050] ;

[0051] ;

[0052] ;

[0053] Where P ij A represents the recommendation priority of customer i for product j, that is, the recommendation weight of the product to customer i; i represents the suitability score of customer i; S j represents the fitness score of product j, that is, the demand intensity or quality score of the product itself; Q j represents the business relevance of product j; H j Indicates the timeliness of product j, that is, whether the product meets the timeliness requirements of the current market; , , Represents the adjustment coefficient, which is used to balance the impact of various factors in the recommendation priority; , Indicates the weight of each factor, that is, the contribution of different factors to the product fit score; represents the market demand intensity of product j; Indicates the quality rating of the product;

[0054] S52. Sort the recommendation priorities, put the products with high priority in front, and get the customer's recommendation list:

[0055] ;

[0056] In the formula, List i represents the recommendation list of customer i; P ijrepresents the recommendation priority of customer i for product j.

[0057] Preferably, the process of generating an identification tag for customer data is as follows:

[0058] S61. Select a random number k∈[1, p-1], calculate the auxiliary coding factor FC=k×G, and define the coordinates of the auxiliary coding factor FC on the elliptic curve E as ( x , y );

[0059] Wherein, G is a base point of the elliptic curve E defined on the finite field GF(q), and the order of the base point G is p;

[0060] S62. Calculate the inspection index Ind: ;

[0061] If Ind=0, re-execute S61~S62;

[0062] Where H is a secure and collision-resistant hash function; r is a random number generated in the initialization phase, r∈(0, p-1); Cdata is the binary encoding form of the customer data;

[0063] S63, generate an identification tag (Ind, x ).

[0064] Preferably, the verification process for verifying the accuracy of customer data is as follows:

[0065] S71, check whether Ind is in the interval [1, p-1], if Ind∈[1, p-1], continue to execute S72, otherwise an error is reported and a safety warning is issued;

[0066] S72, calculate the test factor AF = (Ind) -1 ×H(Cdata) mod p;

[0067] S73. Calculate the point on the elliptic curve: x' , y' )=AF×EF;

[0068] like x' = x , then the customer data Cdata received by the customer portrait generation module is accurate;

[0069] Among them, EF is the data deblocking factor generated in the initialization phase, EF=r×G; r is a random number generated in the initialization phase, r∈(0, p-1).

[0070] The technical solution of the present invention is a business suitability intelligent analysis system combined with customer portrait, which is used to execute the business suitability intelligent analysis method combined with customer portrait mentioned in the claim, including:

[0071] Customer data collection module, used to collect customer data from various channels, including customer behavior data, transaction data and feedback data;

[0072] The customer portrait generation module is used to analyze the collected customer data and generate the customer portrait;

[0073] Business demand modeling module, used to collect business demand data, build business demand models, and generate demand scoring vectors;

[0074] The suitability analysis module is used to evaluate the suitability of customers and various services based on customer profiles and demand scoring vectors, and generate a suitability score;

[0075] The recommendation and decision-making module is used to provide enterprises with decision support for optimizing business, recommending products, and adjusting strategies based on the suitability score;

[0076] The feedback and optimization module is used to continuously optimize user portraits, suitability scores, and intelligent recommendation algorithms based on customer feedback and market changes.

[0077] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects:

[0078] The present invention designs a business adaptability intelligent analysis method and system combined with customer portraits, which improves the accuracy of personalized recommendations, optimizes business decisions, enhances customer loyalty, adjusts recommendation strategies in real time, expands cross-channel recommendations, improves resource utilization efficiency and customer satisfaction, and thus achieves more efficient enterprise operations and customer service:

[0079] (1) Improved accuracy of personalized recommendations: By combining multi-dimensional customer data and generating accurate customer portraits, the matching degree between business needs and customer needs is improved, so that more accurate personalized recommendations can be provided. Customer needs are more accurately predicted and understood, and the relevance and adaptability of recommendations are improved.

[0080] (2) Business decision optimization: Through the suitability analysis module, the suitability of customers to specific business needs can be quantified, providing a basis for enterprises to optimize product recommendations, marketing strategies and service solutions, helping enterprises to provide customized services based on customers' actual needs and behaviors, thereby improving operational efficiency and customer satisfaction.

[0081] (3) Improve customer loyalty: The introduction of a customer loyalty scoring model can help companies identify highly loyal customers and take targeted measures to maintain their relationships, thereby increasing customers’ repurchase rates and willingness to recommend products, thereby enhancing customer stickiness.

[0082] (4) Dynamic adjustment and real-time feedback: The recommendation and decision-making module combines real-time customer behavior data and business demand feedback to quickly adjust the recommendation strategy. By continuously optimizing the recommendation algorithm, it ensures the timeliness and accuracy of personalized recommendations and improves customer experience.

[0083] (5) Cross-channel integration and diversified recommendations: Personalized recommendations through multiple channels can cover different customer groups, expand the reach of recommendations, and improve conversion rates and customer satisfaction;

[0084] (6) Optimizing customer profile generation: Through a secure and efficient customer data transmission method, the customer profile generation module can receive the latest customer data in a timely manner. The accurate data generated based on the verification mechanism can be used to build a more accurate customer profile, which helps to improve the accuracy of subsequent fitness analysis and personalized recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 This is a system architecture diagram of a business adaptability intelligent analysis system combined with customer portraits proposed by the present invention;

[0086] Figure 2 This is a method flow chart of a method for intelligent analysis of business adaptability combined with customer portraits proposed by the present invention. DETAILED DESCRIPTION

[0087] Embodiment 1, as Figure 1 As shown, the present invention proposes a business adaptability intelligent analysis system combined with customer portrait, including a customer data collection module, a customer portrait generation module, a business demand modeling module, a adaptability analysis module, a recommendation and decision module and a feedback and optimization module.

[0088] The customer data collection module is used to collect customer data from various channels, including customer behavior data, transaction data and feedback data;

[0089] It should be noted that channels include but are not limited to websites, social media, and customer service;

[0090] The customer portrait generation module is used to analyze the collected customer data and generate a dynamic portrait of the customer using data processing and machine learning algorithms;

[0091] The business demand modeling module is used to collect business demand data, build business demand models, and generate demand scoring vectors;

[0092] The suitability analysis module is used to evaluate the suitability of customers and various services based on customer profiles and demand scoring vectors, and generate a suitability score;

[0093] The recommendation and decision-making module is used to provide enterprises with decision support for optimizing business, recommending products, and adjusting strategies based on the suitability score;

[0094] The feedback and optimization module is used to continuously optimize user portraits, suitability scores, and intelligent recommendation algorithms based on customer feedback and market changes.

[0095] Embodiment 2, as Figure 2 As shown, the present invention proposes a method for intelligent analysis of business adaptability combined with customer portraits, which is applied to a system for intelligent analysis of business adaptability combined with customer portraits proposed in Example 1, and its specific implementation steps are as follows:

[0096] S1. The customer data collection module collects customer data from various data sources, including but not limited to:

[0097] Basic customer information: including but not limited to the customer's name, ID number, age, gender, occupation, geographic location, and contact information;

[0098] Behavioral data: including but not limited to customers’ clicks, browsing, purchases, and searches on corporate websites, applications, and other platforms, and recording customers’ preferences and interests;

[0099] Transaction data: including but not limited to customers’ order records, payment methods, shopping frequency, spending amount, and returns;

[0100] Social data: including but not limited to customers’ likes, comments, and sharing on social media;

[0101] It should be noted that the data sources include but are not limited to:

[0102] Internal enterprise data: basic customer information and transaction data stored in the customer management system (CRM), sales system, and customer service system;

[0103] Online behavior data: customer behavior data on corporate websites, mobile applications, and social media platforms;

[0104] Social media data: customer interactions, comments, and sharing of information in social networks;

[0105] External cooperation data: customer analysis data from third parties, including but not limited to credit reporting agencies and data service providers.

[0106] S2. The customer profile generation module receives the collected customer data and performs the following operations:

[0107] S21. Based on the customer data, extract the customer's ID number and generate a unique customer identification ID;

[0108] S22. Use the TF-IDF (term frequency–inverse document frequency) model to perform text analysis, automatically extract subject words, calculate the weights of different keywords in the text data, and use keywords with high weights as interest tags;

[0109] S23. Construct a customer-interest tag matrix and generate an interest preference vector using the TF-IDF weighting method:

[0110] ;

[0111] ;

[0112] In the formula, I i represents the interest preference vector of customer i, including the customer's preference for various interest tags; T j represents the interest tag j, i.e. the possible interest point of the customer; w ij represents the interest weight of customer i for interest tag j; tf ij represents the frequency of customer i's behavior on interest tag j; N represents the total number of customers; df j represents the number of customers containing interest tag j, that is, the number of customers who show interest in interest tag j;

[0113] S24. Capture the customer's most recent purchase time, purchase frequency and consumption amount from the customer data, build an RFM model, and generate a purchasing power vector P i =[R i , F i , M i ];

[0114] Among them, P i represents the purchasing power vector of customer i; R i represents the most recent purchase time of customer i, that is, the number of days between the most recent purchase and the current time; F i represents the purchase frequency of customer i in a given time period; M i represents the consumption amount of customer i in a given time period. In this embodiment, the total consumption amount is used;

[0115] S25. Capture the customer's repurchase frequency, total purchase frequency, customer satisfaction, and customer recommendation frequency from the customer data, and calculate the loyalty score L i :

[0116] ;

[0117] Where, L i Re represents the loyalty score of customer i; i It represents the proportion of multiple purchases by customer i in a given time period, that is, the number of repeat purchases / total number of purchases; E i represents the evaluation score, i.e., the average satisfaction level of customer i; Ni represents the recommendation behavior score, i.e., the number of times a customer recommends; α, β, γ represent weight coefficients, i.e., the impact of repurchase rate, evaluation score, and recommendation behavior on loyalty;

[0118] S26. Use cosine similarity to calculate the similarity between customers and find the interest preferences of other customers similar to customer i. If the similarity between customer i and customer j is high, the label weight of customer j can be used to update the interest preference vector of customer i to obtain the adjusted interest preference vector of the target customer I'. i :

[0119] ;

[0120] ;

[0121] In the formula, represents the updated interest preference vector of customer i; λ represents the update coefficient, which is used to control the degree of interest preference update; S(i) represents the set of customers similar to customer i; sim(i,j) represents the similarity between customer i and customer j; R ik and R jk denotes the ratings of customer i and customer j on interest tag k respectively; n denotes the total number of interest tags;

[0122] S27, output image vector C i =[B i , I' i , P i , L i ], that is, the customer profile C of the final customer i is obtained i =[B i , I' i , P i , L i ], build association relationship <ID i , C i > and associate the relationship <ID i , C i >Transfer to the fitness analysis module;

[0123] Among them, B i Represents the basic attributes of customer i, including but not limited to the customer's name, age, gender, occupation, geographic location and contact information; ID i Represents the unique customer identifier of customer i.

[0124] S3, the business demand modeling module collects business demand data and builds a business demand model. The specific implementation process is as follows:

[0125] S31. Collect business demand data, including but not limited to:

[0126] Customer behavior data: mining customers’ potential needs based on customer behaviors including but not limited to customer historical transaction records, click logs, and browsing behaviors, for example, the product categories that customers recently browsed, the number of purchases, and the average order amount;

[0127] Business rule data: including but not limited to product recommendation rules and service provision strategies formulated by the company for different business scenarios or market strategies;

[0128] External environmental data: including but not limited to industry trends, market analysis reports, and seasonal changes;

[0129] Customer feedback data: Obtain clear customer demand information through methods including but not limited to customer evaluation, questionnaires, and customer service interactions;

[0130] S32. Based on the collected business demand data, extract the customer's ID number, generate a unique customer ID, and build a multi-dimensional business demand model. Use feature engineering to convert the demand into quantifiable indicators: divide customer demand into different categories, including but not limited to product demand, service demand, and information demand. Each type of demand will be quantified according to the intensity of customer behavior, and a weighted scoring model will be used to quantify the customer's business needs.

[0131] The demand score vector for each dimension is as follows:

[0132] ;

[0133] Where D i represents the business demand score vector of customer i, that is, the demand intensity of the user in different business demand dimensions; w j represents the weight of business requirement j, which reflects the relative importance of different types of requirements; D ij represents the demand score of customer i under demand dimension j, calculated from behavioral data (including but not limited to purchase frequency and browsing times);

[0134] For example, assuming there are three demand dimensions (product demand, service demand, information demand), the demand scores of customer i under these dimensions are:

[0135] ;

[0136] S33. Building association relationships <ID i , Di > and associate the relationship <ID i , D i >Transfer to the fitness analysis module.

[0137] S4, the suitability analysis module combines business requirements with customer profiles to evaluate the customer's suitability for specific business requirements and obtain a suitability score based on this. The specific implementation process is as follows:

[0138] S41. Based on the relationship <ID i , C i > and related relationships <ID i , D i >, the portrait vector C of any customer i i and the demand score vector D i After matching, the similarity between the customer's portrait vector and the demand score vector is calculated:

[0139] ;

[0140] In the formula, C i represents the portrait vector of customer i; D i represents the business requirement score vector of customer i; ||C i || represents the modulus of the portrait vector of customer i; ||D i || represents the modulus of the business requirement score vector of customer i;

[0141] S42. Generate a fitness score A by combining the similarity between the customer portrait vector and the demand score vector i :

[0142] ;

[0143] ;

[0144] In the formula, A i represents the suitability score of customer i, that is, the degree of match between customer i and specific business requirements; and Represents the adjustment coefficient, which controls the contribution of similarity calculation and other factors (including but not limited to the influence of historical behavior, etc.) to the fitness score; represents an additional function used to enhance the fitness score; represents the historical purchase frequency of a certain type of product by customer i; represents the weighting coefficient, which is used to adjust the impact of historical behavior on the fitness score; represents the number of interactive behaviors of customer i in a given time; represents the weighting coefficient, which is used to control the contribution of activity to the fitness score; Represents the weighting coefficient, which is used to adjust the intensity of time decay; It means that as time goes by, the customer's suitability for a certain demand gradually decreases; represents the attenuation factor, which controls the influence of time distance on the fitness score; t i Indicates the number of days since customer i’s last interaction with a certain type of business demand;

[0145] S43. Build association relationships <ID i , A i > and associate the relationship <ID i , A i >Transfer to recommendation and decision module.

[0146] S5. Recommendation and decision modules obtain association relationships <ID i , A i >, generate personalized recommendations through intelligent recommendation algorithms to ensure that customers obtain maximum business value, while improving customer satisfaction and conversion rate. The specific implementation process is as follows:

[0147] S51. Combine the customer's fit score with other business indicators (including but not limited to product popularity and profit margin) to derive the recommendation priority:

[0148] ;

[0149] ;

[0150] ;

[0151] Where P ij A represents the recommendation priority of customer i for product j, that is, the recommendation weight of the product to customer i; i represents the suitability score of customer i; S j represents the fitness score of product j, that is, the demand intensity or quality score of the product itself; Q j represents the business relevance of product j; H j Indicates the timeliness of product j, that is, whether the product meets the timeliness requirements of the current market; , , Represents the adjustment coefficient, which is used to balance the impact of various factors in the recommendation priority; , Indicates the weight of each factor, that is, the contribution of different factors to the product fit score; represents the market demand strength of product j, which is quantified based on market demand data (including but not limited to product popularity and sales volume) in this embodiment; The quality score of the product is obtained based on customer evaluation, return rate and user feedback indicators in this embodiment;

[0152] S52. Sort the recommendation priorities, put the products with high priority in front, and get the customer's recommendation list. Customers can first see the products or services that best match their needs:

[0153] ;

[0154] In the formula, List i represents the recommendation list of customer i; P ij represents the recommendation priority of customer i for product j;

[0155] S53. According to the recommendation list, the personalized services are recommended to customers through different channels, including but not limited to:

[0156] Online platforms: recommend relevant products or services to customers;

[0157] Email / SMS push: Send personalized promotional information based on recommendation priority;

[0158] Customer service system: recommend relevant products or services based on customer needs.

[0159] S6, Feedback and Optimization Module builds an automated data processing and model update process, using batch processing systems and stream data processing frameworks (Apache Kafka) to process and update customer behavior data, business demand data, and customer feedback data in real time, and then update user portraits, fitness scores, and intelligent recommendation algorithms, continuously improving the accuracy and real-time performance of the system.

[0160] Embodiment 3, the present invention proposes a method for intelligent analysis of business adaptability combined with customer portraits. Compared with Embodiment 2, it also includes a method for efficient transmission of customer data, and its specific implementation steps are as follows:

[0161] S1. The customer portrait generation module implements the following initialization settings:

[0162] Select a random number r∈(0, p-1), calculate the data deblocking factor EF=r×G, save the data deblocking factor EF, and transmit the random number r to the customer data collection module;

[0163] Wherein, G is a base point of the elliptic curve E defined on the finite field GF(q), and the order of the base point G is p.

[0164] S2. The customer data collection module saves the random number r and performs the following operations:

[0165] S21, collecting customer data, encoding the customer data, and obtaining binary string data Cdata;

[0166] S22. Select a random number k∈[1, p-1], calculate the auxiliary coding factor FC=k×G, and define the coordinates of the auxiliary coding factor FC on the elliptic curve E as ( x , y );

[0167] S23. Calculate the inspection index Ind: ;

[0168] If Ind=0, re-execute S22~S23;

[0169] Where H is a secure and collision-resistant hash function;

[0170] S24, generate identification tag (Ind, x ), and {(Ind, x ), Cdata} is transmitted to the customer portrait generation module.

[0171] S3, the customer portrait generation module receives {(Ind, x ), Cdata}, and perform the following operations:

[0172] S31, check whether Ind is in the interval [1, p-1]. If Ind∈[1, p-1], continue to execute S32. Otherwise, an error is reported and a safety warning is issued;

[0173] S32, calculate the test factor AF = (Ind) -1 ×H(Cdata) mod p;

[0174] S33. Calculate the points on the elliptic curve: x' , y' )=AF×EF;

[0175] like x' = x , then the customer data Cdata received by the customer portrait generation module is accurate, which can ensure the accuracy of the customer portrait vector generated based on it.

[0176] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A business suitability intelligent analysis method combined with customer portrait, characterized in that: The specific implementation steps include the following: S1. Collect customer data and generate identification tags for the customer data; The generation process for generating identification tags for customer data is as follows: S61. Select a random number k∈[1, p-1], calculate the auxiliary coding factor FC=k×G, and define the coordinates of the auxiliary coding factor FC on the elliptic curve E as ( x , y ); Wherein, G is a base point of the elliptic curve E defined on the finite field GF(q), and the order of the base point G is p; S62. Calculate the inspection index Ind: ; If Ind=0, re-execute S61~S62; Where H is a secure and collision-resistant hash function; r is a random number generated in the initialization phase, r∈(0, p-1); Cdata is the binary encoding form of the customer data; S63, generate an identification tag (Ind, x ); S2. Verify the accuracy of customer data, and generate a unique customer ID based on the customer data. Use the TF-IDF model to extract interest tags and calculate weights, build a customer-interest tag matrix, generate a customer portrait, and obtain purchase behavior and loyalty scores from customer data. Calculate the similarity between customers, generate the final customer portrait, and build the association relationship <unique customer ID, customer portrait>; The verification process to verify the accuracy of customer data is as follows: S71, check whether Ind is in the interval [1, p-1], if Ind∈[1, p-1], continue to execute S72, otherwise an error is reported and a safety warning is issued; S72, calculate the test factor AF = (Ind) -1 ×H(Cdata) mod p; S73. Calculate the point on the elliptic curve: x' , y' )=AF×EF; like x' = x , then the customer data Cdata received by the customer portrait generation module is accurate; Among them, EF is the data deblocking factor generated in the initialization phase, EF=r×G; r is the random number generated in the initialization phase, r∈(0, p-1); S3. Collect customer demand data, generate a unique customer ID, and build a multi-dimensional business demand model. Use feature engineering to convert demand into quantitative indicators. Use a weighted scoring model to quantify customer business needs. Finally, output a demand scoring vector and build an association relationship <unique customer ID, demand scoring vector>. S4. According to the association relationship <unique customer identifier, customer profile> and the association relationship <unique customer identifier, demand score vector>, the customer's profile vector and the demand score vector are matched to evaluate the customer's suitability for specific business requirements, and a suitability score is obtained accordingly; S5. Based on the fitness score, generate personalized recommendations through the intelligent recommendation algorithm, obtain the recommendation priority, generate a recommendation list, and make personalized recommendations accordingly; S6. Build an automated data processing and model update process, and use batch processing systems and stream data processing frameworks to process and update customer behavior data, business demand data, and customer feedback data in real time, thereby updating user portraits, fitness scores, and intelligent recommendation algorithms.

2. According to claim 1, a method for intelligent analysis of business suitability combined with customer portraits is characterized in that: The final customer portrait generation process is as follows: S21. Use the TF-IDF model to perform text analysis, automatically extract keywords, calculate the weights of different keywords in the text data, and use keywords with high weights as interest tags; S22. Construct a customer-interest tag matrix and generate an interest preference vector using the TF-IDF weighting method: ； ； In the formula, I i represents the interest preference vector of customer i, including the customer's preference for various interest tags; T j represents the interest tag j, i.e. the possible interest point of the customer; w ij represents the interest weight of customer i for interest tag j; tf ij represents the frequency of customer i's behavior on interest tag j; N represents the total number of customers; df j represents the number of customers containing interest tag j, that is, the number of customers who show interest in interest tag j; S23. Capture the customer’s most recent purchase time, purchase frequency, and consumption amount from the customer data, build an RFM model, and generate a purchasing power vector P i =[R i , F i , M i ]; Among them, P i represents the purchasing power vector of customer i; R i represents the most recent purchase time of customer i, that is, the number of days between the most recent purchase and the current time; F i represents the purchase frequency of customer i in a given time period; M i represents the consumption amount of customer i in a given time period; S24. Capture the customer's repurchase frequency, total purchase frequency, customer satisfaction, and customer recommendation frequency from the customer data and calculate the loyalty score L i : ； Where, L i Re represents the loyalty score of customer i; i It represents the proportion of multiple purchases by customer i in a given time period, that is, the number of repeat purchases / total number of purchases; E i represents the evaluation score, i.e., the average satisfaction level of customer i; Ni represents the recommendation behavior score, i.e., the number of times a customer recommends; α, β, γ represent weight coefficients, i.e., the impact of repurchase rate, evaluation score, and recommendation behavior on loyalty; S25. Use cosine similarity to calculate the similarity between customers and find the interest preferences of other customers similar to customer i. If the similarity between customer i and customer j is high, the label weight of customer j can be used to update the interest preference vector of customer i to obtain the interest preference vector of the adjusted target customer I'. i : ； ； In the formula, represents the updated interest preference vector of customer i; λ represents the update coefficient, which is used to control the degree of interest preference update; S(i) represents the set of customers similar to customer i; sim(i,j) represents the similarity between customer i and customer j; R ik and R jk denotes the ratings of customer i and customer j on interest tag k respectively; n denotes the total number of interest tags; S26, output image vector C i =[B i , I' i , P i , L i ], that is, the customer profile C of the final customer i is obtained i =[B i , I' i , P i , L i ]; Among them, B i Represents the basic attributes of customer i.

3. According to claim 1, a method for intelligent analysis of business suitability combined with customer portraits is characterized in that: The construction process of the association relationship <unique customer ID, demand scoring vector> is as follows: S31. Collect business demand data; S32. Based on the collected business demand data, extract the customer's ID number and generate a unique customer identification ID; S33. Build a multi-dimensional business demand model and convert demand into quantifiable indicators through feature engineering: divide customer demand into different categories, quantify each category of demand according to the intensity of customer behavior, and use a weighted scoring model to quantify customer business needs; The demand score vector for each dimension is as follows: ； Where D i represents the business demand score vector of any customer i, that is, the demand intensity of the user in different business demand dimensions; w j represents the weight of business requirement j, which reflects the relative importance of different types of requirements; D ij represents the demand score of customer i under demand dimension j; S34, build association relationship < unique customer ID i , the business demand score vector D of any customer i i > Among them, ID i Represents the unique customer identifier of any customer i.

4. According to claim 1, a method for intelligent analysis of business suitability combined with customer portraits is characterized in that: The process of generating the fitness score is as follows: S41. According to the association relationship <unique customer ID, customer portrait> and the association relationship <unique customer ID, demand scoring vector>, the portrait vector C of any customer i is i and the demand score vector D i Make a match; S42. Based on the portrait vector and the demand score vector of customer i, calculate the similarity between the two: ； In the formula, C i represents the portrait vector of customer i; D i represents the business requirement score vector of customer i; ||C i || represents the modulus of the portrait vector of customer i; ||D i || represents the modulus of the business requirement score vector of customer i; S43. Generate a fitness score A by combining the similarity between the customer portrait vector and the demand score vector i : ； ； In the formula, A i represents the suitability score of customer i, that is, the degree of match between customer i and specific business requirements; and represents the adjustment coefficient, which controls the contribution of similarity calculation and other factors to the fitness score; represents an additional function used to enhance the fitness score; represents the historical purchase frequency of a certain type of product by customer i; represents the weighting coefficient, which is used to adjust the impact of historical behavior on the fitness score; represents the number of interactive behaviors of customer i in a given time; represents the weighting coefficient, which is used to control the contribution of activity to the fitness score; Represents the weighting coefficient, which is used to adjust the intensity of time decay; It means that as time goes by, the customer's suitability for a certain demand gradually decreases; Represents the attenuation factor, which controls the influence of time distance on the fitness score; t i Indicates the number of days since customer i’s last interaction with a certain type of business demand.

5. According to claim 1, a method for intelligent analysis of business suitability combined with customer portraits is characterized in that: The process of generating the recommendation list is as follows: S51. Combine the customer's suitability score with other business indicators to derive the recommendation priority: ； ； ； Where P ij A represents the recommendation priority of customer i for product j, that is, the recommendation weight of the product to customer i; i represents the suitability score of customer i; S j represents the fitness score of product j, that is, the demand intensity or quality score of the product itself; Q j represents the business relevance of product j; H j Indicates the timeliness of product j, that is, whether the product meets the timeliness requirements of the current market; , , Represents the adjustment coefficient, which is used to balance the impact of various factors in the recommendation priority; , Indicates the weight of each factor, that is, the contribution of different factors to the product fit score; represents the market demand intensity of product j; Indicates the quality rating of the product; S52. Sort the recommendation priorities, put the products with high priority in front, and get the customer's recommendation list: ； In the formula, List i represents the recommendation list of customer i; P ij represents the recommendation priority of customer i for product j.

6. A business suitability intelligent analysis system combined with customer portrait, which is used to execute the business suitability intelligent analysis method combined with customer portrait according to any one of claims 1 to 5, characterized in that: include: Customer data collection module, used to collect customer data from various channels, including customer behavior data, transaction data and feedback data; The customer portrait generation module is used to analyze the collected customer data and generate the customer portrait; Business demand modeling module, used to collect business demand data, build business demand models, and generate demand scoring vectors; The suitability analysis module is used to evaluate the suitability of customers and various services based on customer profiles and demand scoring vectors, and generate a suitability score; The recommendation and decision-making module is used to provide enterprises with decision support for optimizing business, recommending products, and adjusting strategies based on the suitability score; The feedback and optimization module is used to continuously optimize user portraits, suitability scores, and intelligent recommendation algorithms based on customer feedback and market changes.

Images