Accurate customer obtaining method based on user behavior analysis

Abstract

The invention discloses an accurate customer obtaining method based on user behavior analysis, and relates to the technical field of cross-border e-commerce, and the technical scheme effectively solves the difficulty of cross-border e-commerce platform data acquisition and integration through a front-end interface and a task scheduler, dynamically collects and processes user behavior data, forms a unified multi-platform fusion data set, and improves the user experience. The data integration difficulty is effectively reduced; meanwhile, through dimensionless processing and a collaborative filtering model, a data island phenomenon is broken, centralized analysis of data of different platforms is realized, and the accuracy of a user portrait is improved; besides, a real-time data feedback mechanism ensures dynamic updating of the potential coefficient of the user, so that the system can timely adjust and accurately obtain a customer list, and the problem of insufficient real-time performance is solved. According to the comprehensive method, the data integration capability is improved, the data island is broken, and the real-time data processing capability is enhanced, so that the customer recommendation and marketing strategies are optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of cross-border e-commerce, and specifically to a precise customer acquisition method based on user behavior analysis. Background Art

[0002] With the rise of cross-border e-commerce platforms, the market scope of enterprises is no longer limited to their own country, and the international market has become a new growth point. However, facing consumers from different countries and regions, how to accurately and efficiently acquire potential customers has become one of the core challenges of cross-border e-commerce; the customer acquisition method of cross-border e-commerce relies on the integration of big data, artificial intelligence, machine learning and multi-channel marketing technology.

[0003] However, although the above customer acquisition methods are effective in the field of cross-border e-commerce, they also face the following technical shortcomings and challenges:

[0004] 1. Difficulty in data acquisition and integration: Cross-border e-commerce involves different countries and regions, and user behavior data is scattered across multiple platforms and channels, making it difficult to acquire and integrate this data. At the same time, privacy protection policies in different regions (such as GDPR) restrict the collection and use of data, making comprehensive analysis of user behavior more complicated.

[0005] 2. Data island phenomenon: Cross-border e-commerce usually uses multiple third-party platforms (such as Amazon and eBay) for transactions. The data of these platforms cannot be easily shared, forming data islands, resulting in one-sided analysis of user behavior and inability to fully and accurately evaluate user shopping habits and preferences.

[0006] 3. Lack of real-time performance: User behavior in cross-border e-commerce is highly dynamic, especially during promotions or special holidays, when user behavior changes rapidly. If the customer acquisition system cannot achieve real-time analysis and response, important business opportunities or user conversion opportunities may be missed. Summary of the invention

[0007] In view of the deficiencies in the prior art, the present invention provides a precise customer acquisition method based on user behavior analysis, which solves the technical shortcomings mentioned in the background technology.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions: a precise customer acquisition method based on user behavior analysis, comprising the following steps:

[0009] S1. Submit user behavior data collection tasks through the front-end interface. The task scheduler allocates collection tasks and collects user behavior data from multiple cross-border e-commerce platforms and user channels.

[0010] S2. Pre-process the user behavior data collected in step S1, and integrate data from different sources to form a multi-platform integrated data set. At the same time, conduct compliance checks on the collected data to ensure that the multi-platform integrated data set complies with the privacy and data protection regulations of various regions.

[0011] S3, extract features from the user behavior data in the data fusion set, and obtain the user activity coefficient Hd, purchase propensity coefficient Gx, interaction frequency coefficient Jp and regional distribution coefficient Db respectively; then establish a user behavior analysis model, and calculate the target customer potential coefficient Kh through the user behavior analysis model, then calculate and obtain the conversion probability coefficient Zg and evaluate it, and finally adjust the weight of the recommendation relevance coefficient Tx according to the evaluation content of the conversion probability coefficient Zg;

[0012] S4. Based on the user feature data extracted in step S3 and the model prediction results, the target customer potential coefficient Kh is compared with the preset customer potential threshold, and a real-time data feedback mechanism is established and combined to dynamically update the customer recommendation list.

[0013] Preferably, the front-end interface is used to receive behavioral data collection tasks submitted by users, wherein the collection tasks include task description, platform type, data format, data scale, and data integration requirement information, and send the collection tasks to the task scheduler. The task scheduler dynamically collects data according to task requirements, including obtaining user behavior data from different cross-border e-commerce platforms and user channels.

[0014] Preferably, the user access logs and transaction data records of the e-commerce platform are combined with third-party analysis tools to monitor and record user behavior data obtained from different cross-border e-commerce platforms and user channels in real time, including user browsing behavior-related data, purchase record-related data, social interaction-related data, and region-related data, and the collected data is cleaned, deduplicated, and standardized, and finally a multi-platform integrated data set is constructed.

[0015] Preferably, the data format is unified through a data preprocessing module, and a user activity coefficient Hd, a purchase propensity coefficient Gx, an interaction frequency coefficient Jp and a regional distribution coefficient Db are generated in the processed multi-platform fusion data set;

[0016] Among them, the user activity coefficient Hd is obtained by extracting the number of browsing times Lcs, interaction frequency Jpl and total browsing time Zll from the multi-platform fusion data set and performing dimensionless processing. The specific calculation formula is:

[0017]

[0018] The purchase propensity coefficient Gx is obtained by extracting the number of purchases Gcs, the average purchase amount Pgm, and the number of active days Hts from the multi-platform fusion data set and performing dimensionless processing. The specific calculation formula is:

[0019]

[0020] Preferably, the interaction frequency coefficient Jp is calculated by extracting the message interaction times Xxj, the comment times Pcs and the total interaction times Zhd in the multi-platform fusion data set and performing dimensionless processing, and the specific calculation formula is:

[0021]

[0022] The regional distribution coefficient Db is calculated by extracting the regional popularity Lxd, average sales performance Pxs and regional total population Dzr from the multi-platform fusion data set and performing dimensionless processing. The specific calculation formula is:

[0023]

[0024] Preferably, the user activity coefficient Hd, purchase tendency coefficient Gx, interaction frequency coefficient Jp and regional distribution coefficient Db are extracted to generate a user feature vector as the input of the user behavior analysis model; then the collaborative filtering recommendation model is adopted and trained, and finally the user behavior analysis model is constructed; at the same time, the user activity coefficient Hd, purchase tendency coefficient Gx, interaction frequency coefficient Jp and regional distribution coefficient Db are extracted and dimensionless processed, and the target customer potential coefficient Kh is calculated by the following formula:

[0025]

[0026] Preferably, relevant data from the multi-platform fusion data set is extracted and dimensionless processed, and then fitted with the target customer potential coefficient Kh to obtain the conversion probability coefficient Zg; the specific formula is:

[0027]

[0028] In the formula, Tjs represents the recommendation acceptance rate, Gfq represents the shopping cart abandonment rate, and Lzh represents the historical conversion rate;

[0029] The preset conversion probability threshold T is compared with the conversion probability coefficient Zg for evaluation. The specific evaluation content is as follows:

[0030] If the conversion probability threshold T ≤ the conversion probability coefficient Zg, then the user's conversion probability is considered to meet the criteria for key recommended users;

[0031] If the conversion probability threshold T> the conversion probability coefficient Zg, it is considered that the conversion probability of the user does not meet the criteria for key recommended users.

[0032] Preferably, according to the evaluation content of the conversion probability coefficient Zg, the conversion probability threshold T and the conversion probability coefficient Zg are fitted to calculate the recommended relevance coefficient Tx, and the recommended relevance coefficient Tx is weighted; the specific content is:

[0033] When the conversion probability coefficient Zg meets the requirements, the calculation formula of the recommended relevance coefficient Tx is as follows:

[0034]

[0035] At this time, the recommendation relevance Tx is increased by the weighted coefficient α, the recommendation strength is adjusted, and more personalized recommendation content is customized;

[0036] When the conversion probability coefficient Zg does not meet the requirements, the calculation formula of the recommended relevance coefficient Tx is as follows

[0037]

[0038] At this time, the recommendation relevance Tx is reduced by the suppression coefficient β, the recommendation strength is adjusted, and only basic recommended content is pushed.

[0039] Preferably, in combination with the real-time data feedback mechanism, the target customer potential coefficient Kh is compared with the preset customer potential threshold P, and the specific evaluation content is:

[0040] If the target customer potential coefficient Kh ≥ the preset customer potential threshold P, it means that the user meets or exceeds the preset potential standard and has marketing conversion potential; at this time, the customer is included in the precise customer acquisition list as the focus of subsequent marketing and recommendation;

[0041] If the target customer potential coefficient Kh is less than the preset customer potential threshold P, it means that the user currently does not meet the preset potential standard. At this time, continue to observe and exclude him from the precise customer acquisition list, and then determine in real time whether he is eligible to enter the precise customer acquisition list.

[0042] Preferably, based on changes in customer behavior data, the dynamic changes of target customer potential coefficient Kh are continuously monitored;

[0043] Among them, if the customer's behavior data changes within a specific time period, including changes in the user activity coefficient Hd, purchase propensity coefficient Gx and interaction frequency coefficient Jp, the system will recalculate and evaluate the target customer potential coefficient Kh in real time, and then dynamically adjust the personnel in the precise customer acquisition list, including adding or reducing personnel.

[0044] The present invention provides a precise customer acquisition method based on user behavior analysis. It has the following beneficial effects:

[0045] (1) This precise customer acquisition method based on user behavior analysis solves the technical problem of data acquisition and integration difficulties. Through the combination of the front-end interface and the task scheduler, it realizes the efficient collection of user behavior data from multiple cross-border e-commerce platforms and user channels. The task scheduler dynamically collects data based on specific task requirements, covering user browsing, purchasing, social interaction and region-related data. In steps S1 and S2, the preprocessing and data integration mechanism enables data from different sources to be processed in a unified format to form a multi-platform fusion data set. This effectively reduces the difficulty of cross-platform data collection and integration, ensures seamless connection of data between different platforms, and improves the system's data acquisition capabilities.

[0046] (2) This precise customer acquisition method based on user behavior analysis breaks the data island phenomenon; through data fusion and dimensionless processing, centralized processing of user behavior data from different platforms and channels is achieved in step S3, breaking the "data island" phenomenon; after aggregating multi-platform data, user activity coefficient Hd, purchase propensity coefficient Gx, interaction frequency coefficient Jp and regional distribution coefficient Db are generated, so that cross-platform data can be interconnected and complement each other; at the same time, collaborative filtering models are used for data training to further enhance the correlation and fusion of different data sources and improve the accuracy of user portraits; this multi-platform data fusion method breaks the limitations of data isolation, enables user behavior analysis to transcend platform barriers and provide a more comprehensive user potential assessment.

[0047] (3) This precise customer acquisition method based on user behavior analysis solves the problem of insufficient real-time performance. Through the real-time data feedback mechanism, especially in step S4, by continuously monitoring the dynamic changes of the user potential coefficient Kh and related indicators, the user potential evaluation is updated in real time. After the system monitors the changes in user behavior, it will immediately recalculate the potential coefficient and dynamically adjust the precise customer acquisition list based on the evaluation results. This feedback and update mechanism based on real-time data solves the problem of insufficient timeliness in traditional data analysis, ensuring that marketing strategies can quickly respond to changes in customer behavior, thereby achieving precise customer recommendations and acquisition. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 The present invention is a flowchart of the steps of a precise customer acquisition method based on user behavior analysis. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] Example 1

[0051] See also Figure 1 , a precise customer acquisition method based on user behavior analysis, comprising the following steps:

[0052] S1. Submit user behavior data collection tasks through the front-end interface. The task scheduler allocates collection tasks and collects user behavior data from multiple cross-border e-commerce platforms and user channels.

[0053] S2. Pre-process the user behavior data collected in step S1, and integrate data from different sources to form a multi-platform integrated data set. At the same time, conduct compliance checks on the collected data to ensure that the multi-platform integrated data set complies with the privacy and data protection regulations of various regions.

[0054] S3, extract features from the user behavior data in the data fusion set, and obtain the user activity coefficient Hd, purchase propensity coefficient Gx, interaction frequency coefficient Jp and regional distribution coefficient Db respectively; then establish a user behavior analysis model, and calculate the target customer potential coefficient Kh through the user behavior analysis model, then calculate and obtain the conversion probability coefficient Zg and evaluate it, and finally adjust the weight of the recommendation relevance coefficient Tx according to the evaluation content of the conversion probability coefficient Zg;

[0055] S4. Based on the user feature data extracted in step S3 and the model prediction results, the target customer potential coefficient Kh is compared with the preset customer potential threshold, and a real-time data feedback mechanism is established and combined to dynamically update the customer recommendation list.

[0056] In this embodiment, the collection and integration of user behavior data of the cross-border e-commerce platform are optimized through the front-end interface and task scheduler, which significantly reduces the difficulty of data acquisition and integration. The compliance check of the multi-platform integrated data set solves the data island phenomenon and ensures compliance with privacy and data protection regulations. In feature extraction and model analysis, the system can accurately calculate the user activity coefficient Hd, purchase propensity coefficient Gx, interaction frequency coefficient Jp and geographical distribution coefficient Db. By dynamically adjusting the target customer potential coefficient Kh and conversion probability coefficient Zg, the real-time data processing capability is effectively improved, and accurate customer recommendation list updates are achieved, which not only improves the accuracy of user portraits, but also enhances the real-time data response capability, thereby optimizing marketing strategies and customer conversion effects.

[0057] Example 2

[0058] The front-end interface is used to receive behavioral data collection tasks submitted by users, where the collection tasks include task description, platform type, data format, data scale, and data integration requirement information, and send the collection tasks to the task scheduler. The task scheduler dynamically collects data according to task requirements, including obtaining user behavior data from different cross-border e-commerce platforms and user channels.

[0059] In this embodiment, through the collaboration of the front-end interface and the task scheduler, the process of behavioral data collection is greatly simplified, ensuring comprehensive coverage of task descriptions, platform types, data formats, data scales and integration requirements, thereby improving the accuracy of data collection and the flexibility of the system; the task scheduler can dynamically schedule data collection according to task requirements, solving the problem of diversified data sources in cross-border e-commerce platforms and user channels, which not only significantly reduces the complexity of data integration, but also effectively avoids the phenomenon of data islands, and improves data integration efficiency and quality.

[0060] Example 3

[0061] Through the user access logs and transaction data records of the e-commerce platform, combined with third-party analysis tools, real-time monitoring and recording of user behavior data obtained from different cross-border e-commerce platforms and user channels, including user browsing behavior-related data, purchase record-related data, social interaction-related data, and region-related data, the collected data is cleaned, deduplicated, and standardized, and finally a multi-platform integrated data set is constructed.

[0062] In this embodiment, by real-time monitoring and recording of e-commerce platform user access logs and transaction data, and combining with third-party analysis tools, the system can comprehensively acquire and process user behavior data from different cross-border e-commerce platforms and user channels; this method not only effectively solves the complexity and data island problems in data integration, but also ensures the high quality and consistency of the data; through the comprehensive collection, cleaning, deduplication and standardization of browsing behavior, purchase records, social interactions and regional data, the system ultimately constructs a multi-platform integrated data set, which greatly improves the accuracy and practicality of data analysis, and provides a reliable data foundation for subsequent user behavior analysis and precision marketing.

[0063] Example 4

[0064] The data preprocessing module is used to unify the data format, and the user activity coefficient Hd, purchase propensity coefficient Gx, interaction frequency coefficient Jp and regional distribution coefficient Db are generated in the processed multi-platform fusion data set;

[0065] Among them, the user activity coefficient Hd is obtained by extracting the number of browsing times Lcs, interaction frequency Jpl and total browsing time Zll from the multi-platform fusion data set and performing dimensionless processing. The specific calculation formula is:

[0066]

[0067] The purchase propensity coefficient Gx is obtained by extracting the number of purchases Gcs, the average purchase amount Pgm, and the number of active days Hts from the multi-platform fusion data set and performing dimensionless processing. The specific calculation formula is:

[0068]

[0069] The interaction frequency coefficient Jp is calculated by extracting the message interaction times Xxj, the comment times Pcs and the total interaction times Zhd in the multi-platform fusion data set and performing dimensionless processing. The specific calculation formula is:

[0070]

[0071] The regional distribution coefficient Db is calculated by extracting the regional popularity Lxd, average sales performance Pxs and regional total population Dzr from the multi-platform fusion data set and performing dimensionless processing. The specific calculation formula is:

[0072]

[0073] The user activity coefficient Hd, purchase tendency coefficient Gx, interaction frequency coefficient Jp and regional distribution coefficient Db are extracted to generate user feature vectors as the input of the user behavior analysis model; then the collaborative filtering recommendation model is adopted and trained, and finally the user behavior analysis model is constructed; at the same time, the user activity coefficient Hd, purchase tendency coefficient Gx, interaction frequency coefficient Jp and regional distribution coefficient Db are extracted and dimensionless, and the target customer potential coefficient Kh is calculated by the following formula:

[0074]

[0075] After extracting relevant data from the multi-platform fusion data set and performing dimensionless processing, it is fitted with the target customer potential coefficient Kh to obtain the conversion probability coefficient Zg; the specific formula is:

[0076]

[0077] In the formula, Tjs represents the recommendation acceptance rate, Gfq represents the shopping cart abandonment rate, and Lzh represents the historical conversion rate;

[0078] The preset conversion probability threshold T is compared with the conversion probability coefficient Zg for evaluation. The specific evaluation content is as follows:

[0079] If the conversion probability threshold T ≤ the conversion probability coefficient Zg, then the user's conversion probability is considered to meet the criteria for key recommended users;

[0080] If the conversion probability threshold T> the conversion probability coefficient Zg, it is considered that the conversion probability of the user does not meet the criteria for key recommended users.

[0081] According to the evaluation content of the conversion probability coefficient Zg, the conversion probability threshold T and the conversion probability coefficient Zg are fitted to calculate the recommended relevance coefficient Tx, and the recommended relevance coefficient Tx is weighted; the specific content is:

[0082] When the conversion probability coefficient Zg meets the requirements, the calculation formula of the recommended relevance coefficient Tx is as follows:

[0083]

[0084] At this time, the recommendation relevance Tx is increased by the weighted coefficient α, the recommendation strength is adjusted, and more personalized recommendation content is customized;

[0085] When the conversion probability coefficient Zg does not meet the requirements, the calculation formula of the recommended relevance coefficient Tx is as follows

[0086]

[0087] At this time, the recommendation relevance Tx is reduced by the suppression coefficient β, the recommendation strength is adjusted, and only basic recommended content is pushed.

[0088] In this embodiment, the generated user activity coefficient Hd, purchase propensity coefficient Gx, interaction frequency coefficient Jp and regional distribution coefficient Db provide accurate input data for the user behavior analysis model; these coefficients are dimensionless and calculated with specific formulas, which can comprehensively reflect the user's activity level, purchase intention, interaction frequency and regional distribution, and effectively solve the complexity problem in data integration and analysis; the generation of user feature vectors and the training of collaborative filtering recommendation models further improve the accuracy of user behavior analysis; the calculation of target customer potential coefficient Kh and conversion probability coefficient Zg and their comparison with preset thresholds ensure accurate user recommendations; the weight adjustment of the recommendation relevance coefficient Tx according to the conversion probability coefficient Zg not only optimizes the personalization of the recommended content, but also improves the pertinence and effectiveness of the recommendation effect, thereby solving the problems of data silos and lack of real-time performance.

[0089] Example 5

[0090] Combined with the real-time data feedback mechanism, the target customer potential coefficient Kh is compared with the preset customer potential threshold P. The specific evaluation content is as follows:

[0091] If the target customer potential coefficient Kh ≥ the preset customer potential threshold P, it means that the user meets or exceeds the preset potential standard and has marketing conversion potential; at this time, the customer is included in the precise customer acquisition list as the focus of subsequent marketing and recommendation;

[0092] If the target customer potential coefficient Kh is less than the preset customer potential threshold P, it means that the user currently does not meet the preset potential standard. At this time, continue to observe and exclude him from the precise customer acquisition list, and then determine in real time whether he is eligible to enter the precise customer acquisition list.

[0093] Based on the changes in customer behavior data, continuously monitor the dynamic changes of target customer potential coefficient Kh;

[0094] Among them, if the customer's behavior data changes within a specific time period, including changes in the user activity coefficient Hd, purchase propensity coefficient Gx and interaction frequency coefficient Jp, the system will recalculate and evaluate the target customer potential coefficient Kh in real time, and then dynamically adjust the personnel in the precise customer acquisition list, including adding or reducing personnel.

[0095] In this embodiment, in combination with the real-time data feedback mechanism, by comparing the target customer potential coefficient Kh with the preset customer potential threshold P, customers who meet the potential standards can be dynamically screened, thereby ensuring the accuracy of the precise customer acquisition list; this mechanism not only solves the static problem in potential customer identification, but also timely adjusts the customer list by real-time monitoring of changes in user activity coefficient Hd, purchase propensity coefficient Gx and interaction frequency coefficient Jp, effectively responding to the challenge of insufficient data real-time performance; through continuous tracking and analysis of customer behavior data, the system can dynamically calculate the target customer potential coefficient Kh, ensure the timeliness and pertinence of marketing and recommendations, thereby improving the efficiency and accuracy of user acquisition, and solving the data island and integration problems in traditional methods.

[0096] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A precise customer acquisition method based on user behavior analysis, characterized by: The following steps are involved: S1. Submit user behavior data collection tasks through the front-end interface. The task scheduler allocates collection tasks and collects user behavior data from multiple cross-border e-commerce platforms and user channels. S2. Pre-process the user behavior data collected in step S1, and integrate data from different sources to form a multi-platform fusion data set, and at the same time perform compliance checks on the collected data; S3. Extract features from the user behavior data in the data fusion set, and obtain the user activity coefficient Hd, purchase propensity coefficient Gx, interaction frequency coefficient Jp, and regional distribution coefficient Db respectively; establish a user behavior analysis model, and calculate the target customer potential coefficient Kh through the user behavior analysis model, calculate and obtain the conversion probability coefficient Zg and evaluate it, and adjust the weight of the recommendation relevance coefficient Tx according to the evaluation content of the conversion probability coefficient Zg; S4. Based on the user feature data extracted in step S3 and the model prediction results, the target customer potential coefficient Kh is compared with the preset customer potential threshold, and a real-time data feedback mechanism is established and combined to dynamically update the customer recommendation list.

2. According to claim 1, a precise customer acquisition method based on user behavior analysis is characterized by: The front-end interface is used to receive behavioral data collection tasks submitted by users, where the collection tasks include task description, platform type, data format, data scale, and data integration requirement information, and send the collection tasks to the task scheduler. The task scheduler dynamically collects data according to task requirements, including obtaining user behavior data from different cross-border e-commerce platforms and user channels.

3. According to claim 2, a precise customer acquisition method based on user behavior analysis is characterized in that: Through the user access logs and transaction data records of the e-commerce platform, combined with third-party analysis tools, real-time monitoring and recording of user behavior data obtained from different cross-border e-commerce platforms and user channels, including user browsing behavior-related data, purchase record-related data, social interaction-related data, and region-related data, the collected data is cleaned, deduplicated, and standardized, and finally a multi-platform integrated data set is constructed.

4. According to claim 3, a precise customer acquisition method based on user behavior analysis is characterized in that: The data preprocessing module is used to unify the data format, and the user activity coefficient Hd, purchase propensity coefficient Gx, interaction frequency coefficient Jp and regional distribution coefficient Db are generated in the processed multi-platform fusion data set; Among them, the user activity coefficient Hd is obtained by extracting the number of browsing times Lcs, interaction frequency Jpl and total browsing time Zll from the multi-platform fusion data set and performing dimensionless processing. The specific calculation formula is: The purchase propensity coefficient Gx is obtained by extracting the number of purchases Gcs, the average purchase amount Pgm, and the number of active days Hts from the multi-platform fusion data set and performing dimensionless processing. The specific calculation formula is:

5. According to claim 4, a precise customer acquisition method based on user behavior analysis is characterized in that: The interaction frequency coefficient Jp is calculated by extracting the message interaction times Xxj, the comment times Pcs and the total interaction times Zhd in the multi-platform fusion data set and performing dimensionless processing. The specific calculation formula is: The regional distribution coefficient Db is calculated by extracting the regional popularity Lxd, average sales performance Pxs and regional total population Dzr from the multi-platform fusion data set and performing dimensionless processing. The specific calculation formula is:

6. The accurate customer acquisition method based on user behavior analysis according to claim 5, characterized in that: The user activity coefficient Hd, purchase tendency coefficient Gx, interaction frequency coefficient Jp and regional distribution coefficient Db are extracted to generate user feature vectors as the input of the user behavior analysis model; then the collaborative filtering recommendation model is adopted and trained, and finally the user behavior analysis model is constructed; at the same time, the user activity coefficient Hd, purchase tendency coefficient Gx, interaction frequency coefficient Jp and regional distribution coefficient Db are extracted and dimensionless, and the target customer potential coefficient Kh is calculated by the following formula:

7. The accurate customer acquisition method based on user behavior analysis according to claim 6, characterized in that: After extracting relevant data from the multi-platform fusion data set and performing dimensionless processing, it is fitted with the target customer potential coefficient Kh to obtain the conversion probability coefficient Zg; The specific formula is: In the formula, Tjs represents the recommendation acceptance rate, Gfq represents the shopping cart abandonment rate, and Lzh represents the historical conversion rate; The preset conversion probability threshold T is compared with the conversion probability coefficient Zg for evaluation. The specific evaluation content is as follows: If the conversion probability threshold T ≤ the conversion probability coefficient Zg, then the user's conversion probability is considered to meet the criteria for key recommended users; If the conversion probability threshold T> the conversion probability coefficient Zg, it is considered that the conversion probability of the user does not meet the criteria for key recommended users.

8. The accurate customer acquisition method based on user behavior analysis according to claim 7 is characterized by: According to the evaluation content of the conversion probability coefficient Zg, the conversion probability threshold T and the conversion probability coefficient Zg are fitted to calculate the recommended relevance coefficient Tx, and the recommended relevance coefficient Tx is weighted; the specific content is: When the conversion probability coefficient Zg meets the requirements, the calculation formula of the recommended relevance coefficient Tx is as follows: At this time, the recommendation relevance Tx is increased by the weighted coefficient α, the recommendation strength is adjusted, and more personalized recommendation content is customized; When the conversion probability coefficient Zg does not meet the requirements, the calculation formula of the recommended relevance coefficient Tx is as follows At this time, the recommendation relevance Tx is reduced by the suppression coefficient β, the recommendation strength is adjusted, and only basic recommended content is pushed.

9. The accurate customer acquisition method based on user behavior analysis according to claim 8, characterized in that: Combined with the real-time data feedback mechanism, the target customer potential coefficient Kh is compared with the preset customer potential threshold P. The specific evaluation content is as follows: If the target customer potential coefficient Kh ≥ the preset customer potential threshold P, it means that the user meets or exceeds the preset potential standard and has marketing conversion potential; at this time, the customer is included in the precise customer acquisition list as the focus of subsequent marketing and recommendation; If the target customer potential coefficient Kh is less than the preset customer potential threshold P, it means that the user currently does not meet the preset potential standard. At this time, continue to observe and exclude him from the precise customer acquisition list, and then determine in real time whether he is eligible to enter the precise customer acquisition list.

10. The accurate customer acquisition method based on user behavior analysis according to claim 9, characterized in that: Based on the changes in customer behavior data, continuously monitor the dynamic changes of target customer potential coefficient Kh; Among them, if the customer's behavior data changes within a specific time period, including changes in the user activity coefficient Hd, purchase propensity coefficient Gx and interaction frequency coefficient Jp, the system will recalculate and evaluate the target customer potential coefficient Kh in real time, and then dynamically adjust the personnel in the precise customer acquisition list, including adding or reducing personnel.

Images