E-commerce User Preference Analysis System

Abstract

The present invention discloses an e-commerce user preference analysis system, which relates to the technical field of business data processing, including a consumption module for generating user consumption tags based on user basic information and user behavior data; a social module for reviewing user interests reflected in other users' social influence range based on user behavior data and user social data; an interest module for generating personal-social interest tags, a portrait module for constructing a multi-layered user portrait based on user consumption tags, personal-social interest tags, hot commodity tags, and unpopular commodity tags; and a recommendation module for generating a personalized recommendation function through a user portrait. A plurality of tags are integrated to obtain a multi-layered and three-dimensional user portrait, which improves the accuracy of portrait construction and description and provides a stable foundation for the efficient operation of e-commerce.

Description

Technical Field

[0001] The present application relates to the technical field of business data processing, and more specifically, to an e-commerce user preference analysis system. Background Art

[0002] The background technology of e-commerce user preference analysis involves in-depth mining and analysis of user data to better understand users' purchasing habits, interest preferences and behavior patterns. Build a portrait of each user in advance and integrate multi-dimensional content into it. The technologies used include data mining, machine learning, natural language processing, social network analysis and multi-label learning. Through these technologies, e-commerce platforms can understand users more accurately, provide personalized services, and improve user satisfaction and conversion rates.

[0003] In the prior art, when constructing user portraits, only the relationship between users and products is often considered to form the portrait. With the popularity of online shopping, existing e-commerce is becoming more and more open, and users have formed a social circle on the e-commerce platform. This social situation will affect the accuracy of user portraits and must be taken into account in order to obtain a more accurate user preference analysis result.

[0004] Therefore, how to take users’ social situations into consideration in user portraits to improve the accuracy of portraits is a technical problem that needs to be solved. Summary of the invention

[0005] The present invention provides an e-commerce user preference analysis system to solve the technical problem that the user's social situation is not considered in the prior art, resulting in low portrait accuracy. The system includes:

[0006] The consumption module is used to collect basic user information, user behavior data, user social data and product data of the e-commerce center, and generate user consumption tags based on basic user information and user behavior data;

[0007] The social module is used to construct a user social network map based on the user's social data, and determine the social influence range of the user in the user social network map through the user's consumption tags, and review the user interests reflected in the social influence range of other users based on the user's behavior data and user social data, and obtain a social user interest list after integration;

[0008] The interest module is used to determine the personal interests of each user, obtain the personal interests of the user under social influence based on the personal interests of the user and the social user interest list, and generate personal-social interest tags, and generate popular product tags and unpopular product tags through user behavior data, user social data and product data;

[0009] The portrait module is used to build a multi-layered user portrait based on user consumption tags, personal-social interest tags, popular product tags, and unpopular product tags;

[0010] The recommendation module is used to generate personalized recommendation functions based on user portraits to analyze user preferences.

[0011] In some embodiments of the present application, the consumption module is used to:

[0012] User behavior data includes user purchase records, based on which the number of purchases and purchase prices are obtained, and the average single price is obtained through the number of purchases and purchase prices;

[0013] Generate a price curve of price changes over time by using the purchase times and purchase prices in chronological order, generate a horizontal line based on the average of single prices, and insert the horizontal line into the price curve, dividing the price curve into an upper curve area and a lower curve area through the horizontal line;

[0014] Filter out the records of users purchasing the same product at least twice from the product purchase records, and mark the purchase records in the upper curve area and the lower curve area as key points, distinguish the key point types, each key point type represents the product type of the same product purchased at least twice, and preset the effective time mapped to each product type;

[0015] Determine the number of key points under the same key point type and the time span between adjacent key points, take the key point with a later time among the adjacent key points whose time span exceeds the effective time as the second key point, and the other key points as the first key points, so as to obtain all the first key points and all the second key points under the same key point type, and perform weighted summation of all the first key points and all the second key points under the same key point type to obtain the importance score of the key point type;

[0016] Determine the maximum importance score of the key point type in the upper curve area and the lower curve area, and extend the horizontal line up and down according to the median of the difference between the price corresponding to the key point type with the maximum importance score and the horizontal line to obtain the average price range;

[0017] The price corresponding to the key point type with the maximum importance score between the maximum value of the upper curve area and the average price range is taken as the upward price;

[0018] The consumption level is defined according to the average price range and the upward price. The user consumption labels are divided into different categories according to the different ranges of the consumption level. The user consumption label is a combination of the average price range and the upward price.

[0019] In some embodiments of the present application, the social module is used to:

[0020] A social network graph between users is constructed with users as nodes, social relationships as edges, the strength of social relationships as the length of edges, and the type of social relationships as the weight of edges. The users themselves in the social network graph are recorded as primary users, and other users are recorded as secondary users. A comprehensive influence threshold is determined by the number of secondary users in the social network graph. The comprehensive influence of each secondary user on the primary user is calculated, and the secondary users whose comprehensive influence exceeds the comprehensive influence threshold are retained to obtain the initial social influence range of the primary user.

[0021] Auxiliary users are screened according to their consumption tags within the initial social influence range of the primary user. If the consumption tags of the primary user and the auxiliary user are in the same category, or the upward price of the primary user is within the average price range of the auxiliary users, these auxiliary users are retained and other auxiliary users are eliminated to obtain the social influence range of the user.

[0022] In some embodiments of the present application, the social module is used to:

[0023] User behavior data also includes user consumption data, user browsing data, and user evaluation data;

[0024] The interest tags of auxiliary users are determined through user social data, and the interest tags of auxiliary users are verified and adjusted based on user consumption data, user browsing data and user evaluation data, so as to determine the interest tags of each auxiliary user and build a social user interest list.

[0025] In some embodiments of the present application, the system further includes a comparison module for:

[0026] Before obtaining the user's personal interests under social influence according to the user's personal interests and the social user interest list, finding other user's personal interests similar to the user's personal interests, and adjusting the user's personal interests according to the other user's personal interests, including:

[0027] Generate each user's personal interest tag based on user consumption data, user browsing data and user evaluation data, calculate the similarity between other users' personal interest tags and the main user's personal interest tags, and find other users with similar interests;

[0028] If other users with similar interests are in the social network graph of the main user, the similar interests of the other users are regarded as explicit similar interests;

[0029] Otherwise, the similar interests of the other users are regarded as invisible similar interests;

[0030] Different adjustment weights are given to explicit similar interests and implicit similar interests to adjust the personal interest tags of the main user.

[0031] In some embodiments of the present application, the interest module is used to:

[0032] According to the user's personal interests and the social user interest list, the user's personal interests under social influence are obtained, and personal-social interest tags are generated

[0033] Calculate the similarity between the adjusted personal interest tags of the main user and the interest tags in the social user's interest list, take the interest tags in the social user's interest list whose similarity exceeds the similarity threshold as similar social interest tags, and correct the personal interest tags of the main user by using the similar social interest tags, thereby generating personal-social interest tags;

[0034] ;

[0035] in, is the value or weight of the i-th personal interest tag of the main user after correction of similar social interest tags, is the value or weight of the i-th personal interest tag of the main user after adjustment of explicit similar interests and implicit similar interests, exp is an exponential function, n is the number of similar social interest tags whose similarity with the i-th personal interest tag of the main user exceeds the similarity threshold, is the social weight of the jth similar social interest tag similar to the i-th personal interest tag, is the similarity between the i-th personal interest tag and the j-th similar social interest tag, is the similarity threshold, and k is a preset constant.

[0036] In some embodiments of the present application, the interest module is used to:

[0037] The first, second and third most popular indicators are selected through user behavior data, user social data and product data respectively;

[0038] The first hot index, the second hot index and the third hot index are integrated and weighted to obtain a comprehensive hot index, and the hot product label and the unpopular product label of the product are generated according to the size of the comprehensive hot index.

[0039] In some embodiments of the present application, the portrait module is used to:

[0040] User consumption tags, personal-social interest tags, hot product tags and unpopular product tags are standardized, and weights are assigned to user consumption tags, personal-social interest tags, hot product tags and unpopular product tags. User consumption tags, personal-social interest tags, hot product tags and unpopular product tags are multi-label fused to build a multi-level user portrait.

[0041] By applying the above technical solutions, the consumption module is used to collect the basic information, user behavior data, user social data and commodity data of the e-commerce center, and generate user consumption tags according to the basic information and user behavior data of the user; the social module is used to build a user social network map according to the user social data, and determine the social influence range of the user in the user social network map through the user consumption tag, and review the user interests reflected in the social influence range of other users based on the user behavior data and the user social data, and obtain the social user interest list after integration; the interest module is used to determine the user's personal interests of each user, obtain the user's personal interests under social influence according to the user's personal interests and the social user interest list, and generate personal-social interest tags, and generate popular commodity tags and unpopular commodity tags through user behavior data, user social data and commodity data; the portrait module is used to build a multi-layer labeled user portrait according to the user consumption tag, personal-social interest tag, popular commodity tag and unpopular commodity tag; the recommendation module is used to generate a personalized recommendation function through the user portrait to analyze the user preference. This application generates user consumption tags through user basic information and user behavior data, adaptively and dynamically divides the user consumption level and the degree of upward price, and provides a stable foundation for subsequent social influence. Based on user behavior data and user social data, the user interests reflected in the social lives of other users within the social influence range are reviewed. The user interests shown by other users in the social circle are considered, and the social user interests that may affect the user are counted. The user's personal interests under social influence are obtained based on the user's personal interests and the social user interest list. The user's personal interests are adjusted according to the user interests shown by other people on the social network. The social situation is considered to ensure the reliability and accuracy of the user's personal interests. A multi-layered user portrait is constructed based on user consumption tags, personal-social interest tags, popular product tags, and unpopular product tags. Multiple tags are integrated to obtain a multi-level and three-dimensional user portrait, which improves the accuracy of portrait construction and description, and provides a stable foundation for the efficient operation of e-commerce. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0043] Figure 1 A schematic diagram of the structure of an e-commerce user preference analysis system proposed in an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0045] The present application embodiment provides an e-commerce user preference analysis system, which includes the following contents:

[0046] The consumption module is used to collect basic user information, user behavior data, user social data and product data of the e-commerce center, and generate user consumption tags based on basic user information and user behavior data.

[0047] In this embodiment, the basic information of the user includes information such as ID and gender, and the user behavior data includes consumption data, evaluation, interaction, web browsing, etc. The user consumption label describes the user's average consumption level and the upward price level. The upward price level refers to how much the user can exceed the average consumption level to make high consumption. Here, the maximum purchase price is not used as the upward price, but is determined by multiple key points in the curve.

[0048] In some embodiments of the present application, the consumption module is used to:

[0049] User behavior data includes user purchase records, based on which the number of purchases and purchase prices are obtained, and the average single price is obtained through the number of purchases and purchase prices;

[0050] Generate a price curve of price changes over time by using the purchase times and purchase prices in chronological order, generate a horizontal line based on the average of single prices, and insert the horizontal line into the price curve, dividing the price curve into an upper curve area and a lower curve area through the horizontal line;

[0051] Filter out the records of users purchasing the same product at least twice from the product purchase records, and mark the purchase records in the upper curve area and the lower curve area as key points, distinguish the key point types, each key point type represents the product type of the same product purchased at least twice, and preset the effective time mapped to each product type;

[0052] Determine the number of key points under the same key point type and the time span between adjacent key points, take the key point with a later time among the adjacent key points whose time span exceeds the effective time as the second key point, and the other key points as the first key points, so as to obtain all the first key points and all the second key points under the same key point type, and perform weighted summation of all the first key points and all the second key points under the same key point type to obtain the importance score of the key point type;

[0053] Determine the maximum importance score of the key point type in the upper curve area and the lower curve area, and extend the horizontal line up and down according to the median of the difference between the price corresponding to the key point type with the maximum importance score and the horizontal line to obtain the average price range;

[0054] The price corresponding to the key point type with the maximum importance score between the maximum value of the upper curve area and the average price range is taken as the upward price;

[0055] The consumption level is defined according to the average price range and the upward price. The user consumption labels are divided into different categories according to the different ranges of the consumption level. The user consumption label is a combination of the average price range and the upward price.

[0056] In this embodiment, at least two purchases of the same product are selected from the user's purchase records, and key points are determined on the curve from multiple purchases of the same product. The preset effective time mapped to each product type refers to the effective length of the time interval between multiple purchases of the same product determined by historical data. For example, if a product is purchased again after a period of time, it can effectively illustrate the cost-effectiveness of the product, especially the price that is most acceptable to the user. If it exceeds this time, the strength of this explanation will be reduced, so different products have different effective times.

[0057] In this embodiment, the number of key points under the same key point type and the time span between adjacent key points are determined, and the key point with a later time among the adjacent key points whose time span exceeds the effective time is taken as the second key point, and the other key points are taken as the first key point. The same key point type refers to the same commodity or the same price, so multiple purchases are represented as multiple key points on the curve, with the same ordinate price and different abscissa time. The key point with a time span greater than the effective time is taken as the second key point, and the first key point is the purchase record within the effective time. The weighted sum of all first key points and all second key points under the same key point type is obtained to obtain the importance score of the key point type, where the weight of the first key point is greater than the weight of the second key point. Here, the importance score of the key point type can be determined by measuring the number of key points and the time span, and this score is used to describe the user's preference for price.

[0058] In this embodiment, the maximum importance score of the key point type in the upper curve area and the lower curve area is determined, and the horizontal line is extended up and down according to the median of the difference between the price corresponding to the key point type with the maximum importance score and the horizontal line to obtain the average price range. If only the original average price is used as the average value to measure the user's consumption level, the user's preference for price may be ignored. Therefore, the original average value is extended here by the key point type. The horizontal line is extended up and down to the median of the maximum score on the upper curve area and the lower curve area and the horizontal line. The price corresponding to the key point type with the maximum importance score between the maximum value of the upper curve area and the average price range is used as the upward price, which means that the upward price is the price corresponding to the key point type with the maximum importance score between the price maximum and the upper end point of the price maximum-average price range. The upward price is the degree of high consumption that the user can accept.

[0059] The social module is used to build a user social network graph based on user social data, and determine the scope of social influence on users in the user social network graph through user consumption tags. Based on user behavior data and user social data, the user interests reflected in the social lives of other users within the scope of social influence are reviewed, and a social user interest list is obtained after integration.

[0060] In this embodiment, user social data includes the user's social interaction behaviors, such as following and being followed, liking and being liked, commenting and being commented on, etc. The social influence range is filtered through user consumption tags, and the user interests of other users in the range reflected in social interaction are determined.

[0061] In some embodiments of the present application, the social module is used to:

[0062] A social network graph between users is constructed with users as nodes, social relationships as edges, the strength of social relationships as the length of edges, and the type of social relationships as the weight of edges. The users themselves in the social network graph are recorded as primary users, and other users are recorded as secondary users. A comprehensive influence threshold is determined by the number of secondary users in the social network graph. The comprehensive influence of each secondary user on the primary user is calculated, and the secondary users whose comprehensive influence exceeds the comprehensive influence threshold are retained to obtain the initial social influence range of the primary user.

[0063] Auxiliary users are screened according to their consumption tags within the initial social influence range of the primary user. If the consumption tags of the primary user and the auxiliary user are in the same category, or the upward price of the primary user is within the average price range of the auxiliary users, these auxiliary users are retained and other auxiliary users are eliminated to obtain the social influence range of the user.

[0064] In this embodiment, a comprehensive influence threshold is determined by the number of auxiliary users in the social network graph. The more auxiliary users there are, the greater the comprehensive influence threshold is. In order to filter out some users with social influence, the more auxiliary users there are, generally speaking, the more auxiliary users are filtered out. In order to improve the accuracy of social influence, this correspondence can be obtained through historical experience or mathematical models. Comprehensive influence can be based on the total weight of the user or the weighted influence (edge ​​weight and length in the graph).

[0065] In this embodiment, if the consumption tags of the main user and the auxiliary user are of the same category, or the upward price of the main user is within the average price range of the auxiliary user, these auxiliary users will be retained, and the users within the range will be screened again to remove the influence of social users with too large differences in consumption levels. For example, the main user a and the auxiliary user b, a's average price range is 200-300, the upward price is 450, and b's average price range is 600-1000, the upward price is 2000. Obviously, a's maximum acceptable price of 450 is not within b's average price range, so b's interest preference is likely to be beyond a's reach, and b will be eliminated.

[0066] In some embodiments of the present application, the social module is used to:

[0067] User behavior data also includes user consumption data, user browsing data, and user evaluation data;

[0068] The interest tags of auxiliary users are determined through user social data, and the interest tags of auxiliary users are verified and adjusted based on user consumption data, user browsing data and user evaluation data, so as to determine the interest tags of each auxiliary user and build a social user interest list.

[0069] In this embodiment, only the social performance of the auxiliary user is considered for interest confirmation, because the main user is more easily influenced by the interests that can be shown in social situations, which is an explicit interest influence. The interest tags of the auxiliary user are verified and adjusted based on the user consumption data, user browsing data and user evaluation data, and these data are utilized using text analysis techniques such as keyword extraction and topic modeling to verify the interest tags (values ​​or weights) of the auxiliary user.

[0070] In some embodiments of the present application, the system further includes a comparison module for:

[0071] Before obtaining the user's personal interests under social influence according to the user's personal interests and the social user interest list, finding other user's personal interests similar to the user's personal interests, and adjusting the user's personal interests according to the other user's personal interests, including:

[0072] Generate each user's personal interest tag based on user consumption data, user browsing data and user evaluation data, calculate the similarity between other users' personal interest tags and the main user's personal interest tags, and find other users with similar interests;

[0073] If other users with similar interests are in the social network graph of the main user, the similar interests of the other users are regarded as explicit similar interests;

[0074] Otherwise, the similar interests of the other users are regarded as invisible similar interests;

[0075] Different adjustment weights are given to explicit similar interests and implicit similar interests to adjust the personal interest tags of the main user.

[0076] In this embodiment, the similarity between the personal interest tags of other users and the main user is calculated. Here, the cosine similarity can be used to determine the explicit similar interests and the implicit similar interests. The former has a larger adjustment weight, while the latter has a smaller adjustment weight.

[0077] The interest module is used to determine the personal interests of each user, obtain the user's personal interests under social influence based on the user's personal interests and the social user interest list, and generate personal-social interest tags, and generate popular product tags and unpopular product tags through user behavior data, user social data and product data.

[0078] In this embodiment, personal-social interest tags are generated based on the interest list of social users, that is, personal interest tags influenced by the interests expressed by social users. Popular product tags and unpopular product tags are generated through user behavior data, user social data and product data, and products are divided into popular products and unpopular products. Unpopular products can better reflect user preferences, so as to make user portraits more accurate.

[0079] In some embodiments of the present application, the interest module is used to:

[0080] According to the user's personal interests and the social user interest list, the user's personal interests under social influence are obtained, and personal-social interest tags are generated

[0081] Calculate the similarity between the adjusted personal interest tags of the main user and the interest tags in the social user's interest list, take the interest tags in the social user's interest list whose similarity exceeds the similarity threshold as similar social interest tags, and correct the personal interest tags of the main user by using the similar social interest tags, thereby generating personal-social interest tags;

[0082] ;

[0083] in, is the value or weight of the i-th personal interest tag of the main user after correction of similar social interest tags, is the value or weight of the i-th personal interest tag of the main user after adjustment of explicit similar interests and implicit similar interests, exp is an exponential function, n is the number of similar social interest tags whose similarity with the i-th personal interest tag of the main user exceeds the similarity threshold, is the social weight of the jth similar social interest tag similar to the i-th personal interest tag, is the similarity between the i-th personal interest tag and the j-th similar social interest tag, is the similarity threshold, and k is a preset constant.

[0084] In this embodiment, the personal interest tag value or weight is modified by the difference between the similarity and the similarity threshold. The value range is between 0.7-1.26.

[0085] In some embodiments of the present application, the interest module is used to:

[0086] The first, second and third most popular indicators are selected through user behavior data, user social data and product data respectively;

[0087] The first hot index, the second hot index and the third hot index are integrated and weighted to obtain a comprehensive hot index, and the hot product label and the unpopular product label of the product are generated according to the size of the comprehensive hot index.

[0088] In this embodiment, the first popular indicator (purchase record, browsing history, quantification of favorites content, etc.), the second popular indicator (friends’ discussion popularity, friends’ recommendation degree, etc.) and the third popular indicator (product sales volume, click-through rate, conversion rate, etc.) are respectively screened out through user behavior data, user social data and product data, and the three popular indicators are integrated into a comprehensive popular indicator. The larger this indicator is, the more popular the product is.

[0089] The portrait module is used to build a multi-layered user portrait based on user consumption tags, personal-social interest tags, popular product tags, and unpopular product tags.

[0090] In some embodiments of the present application, the portrait module is used to:

[0091] User consumption tags, personal-social interest tags, hot product tags and unpopular product tags are standardized, and weights are assigned to user consumption tags, personal-social interest tags, hot product tags and unpopular product tags. User consumption tags, personal-social interest tags, hot product tags and unpopular product tags are multi-label fused to build a multi-level user portrait.

[0092] In this embodiment, different labels are standardized to ensure that they can be compared in terms of dimensions. A weight is assigned to each label to reflect its importance in the user portrait. A fusion algorithm, such as a multi-label learning algorithm (such as One-vs-Rest, Label Powerset, Label Ranking, etc.), is used to fuse multiple labels.

[0093] It should be noted that the present application does not limit the specific multi-label fusion mode or method, which is not the focus of protection of the present application. Other fusion modes that can reflect the idea of ​​the present application are all acceptable, and the personalized recommendation function is the same.

[0094] The recommendation module is used to generate personalized recommendation functions based on user portraits to analyze user preferences.

[0095] In this embodiment, the user portrait is analyzed, the user's interests and preferences are identified, and a suitable recommendation algorithm is selected, such as collaborative filtering, content-based recommendation, hybrid recommendation, etc. The features and preferences in the user portrait are used to train the recommendation model to obtain a personalized recommendation function.

[0096] It should be noted that the corresponding relationship or mapping relationship involved in the above content can be obtained through historical experience or through mathematical models.

[0097] By applying the above technical solutions, the consumption module is used to collect the basic information, user behavior data, user social data and commodity data of the e-commerce center, and generate user consumption tags according to the basic information and user behavior data of the user; the social module is used to build a user social network map according to the user social data, and determine the social influence range of the user in the user social network map through the user consumption tag, and review the user interests reflected in the social influence range of other users based on the user behavior data and the user social data, and obtain the social user interest list after integration; the interest module is used to determine the user's personal interests of each user, obtain the user's personal interests under social influence according to the user's personal interests and the social user interest list, and generate personal-social interest tags, and generate popular commodity tags and unpopular commodity tags through user behavior data, user social data and commodity data; the portrait module is used to build a multi-layer labeled user portrait according to the user consumption tag, personal-social interest tag, popular commodity tag and unpopular commodity tag; the recommendation module is used to generate a personalized recommendation function through the user portrait to analyze the user preference. This application generates user consumption tags through user basic information and user behavior data, adaptively and dynamically divides the user consumption level and the degree of upward price, and provides a stable foundation for subsequent social influence. Based on user behavior data and user social data, the user interests reflected in the social lives of other users within the social influence range are reviewed. The user interests shown by other users in the social circle are considered, and the social user interests that may affect the user are counted. The user's personal interests under social influence are obtained based on the user's personal interests and the social user interest list. The user's personal interests are adjusted according to the user interests shown by other people on the social network. The social situation is considered to ensure the reliability and accuracy of the user's personal interests. A multi-layered user portrait is constructed based on user consumption tags, personal-social interest tags, popular product tags, and unpopular product tags. Multiple tags are integrated to obtain a multi-level and three-dimensional user portrait, which improves the accuracy of portrait construction and description, and provides a stable foundation for the efficient operation of e-commerce.

[0098] Those skilled in the art will appreciate that the modules in the system in the implementation scenario can be distributed in the system of the implementation scenario according to the implementation scenario description, or can be changed accordingly and located in one or more systems different from the implementation scenario. The modules in the above implementation scenario can be combined into one module, or can be further split into multiple sub-modules.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An e-commerce user preference analysis system, characterized in that: include: The consumption module is used to collect basic user information, user behavior data, user social data and product data of the e-commerce center, and generate user consumption tags based on basic user information and user behavior data; The social module is used to construct a user social network map based on the user's social data, and determine the social influence range of the user in the user social network map through the user's consumption tags, and review the user interests reflected in the social influence range of other users based on the user's behavior data and user social data, and obtain a social user interest list after integration; The interest module is used to determine the personal interests of each user, obtain the personal interests of the user under social influence based on the personal interests of the user and the social user interest list, and generate personal-social interest tags, and generate popular product tags and unpopular product tags through user behavior data, user social data and product data; The portrait module is used to build a multi-layered user portrait based on user consumption tags, personal-social interest tags, popular product tags, and unpopular product tags; The recommendation module is used to generate personalized recommendation functions based on user portraits to analyze user preferences; Wherein, the consumption module is used for: User behavior data includes user purchase records, based on which the number of purchases and purchase prices are obtained, and the average single price is obtained through the number of purchases and purchase prices; Generate a price curve of price changes over time by using the purchase times and purchase prices in chronological order, generate a horizontal line based on the average of single prices, and insert the horizontal line into the price curve, dividing the price curve into an upper curve area and a lower curve area through the horizontal line; Filter out the records of users purchasing the same product at least twice from the purchase records of the product, and mark the purchase records in the upper curve area and the lower curve area as key points, distinguish the key point types, each key point type represents the product type of the same product purchased at least twice, and preset the effective time mapped to each product type; Determine the number of key points under the same key point type and the time span between adjacent key points, take the key point with a later time among the adjacent key points whose time span exceeds the effective time as the second key point, and the other key points as the first key points, so as to obtain all the first key points and all the second key points under the same key point type, and perform weighted summation of all the first key points and all the second key points under the same key point type to obtain the importance score of the key point type; Determine the maximum importance score of the key point type in the upper curve area and the lower curve area, and extend the horizontal line up and down according to the median of the difference between the price corresponding to the key point type with the maximum importance score and the horizontal line to obtain the average price range; The price corresponding to the key point type with the maximum importance score between the maximum value of the upper curve area and the average price range is taken as the upward price; The consumption level is defined according to the average price range and the upward price. The user consumption labels are divided into different categories according to the different ranges of the consumption level. The user consumption label is a combination of the average price range and the upward price.

2. The e-commerce user preference analysis system according to claim 1, characterized in that: The social module is used to: A social network graph between users is constructed with users as nodes, social relationships as edges, the strength of social relationships as the length of edges, and the type of social relationships as the weight of edges. The users themselves in the social network graph are recorded as primary users, and other users are recorded as secondary users. A comprehensive influence threshold is determined by the number of secondary users in the social network graph. The comprehensive influence of each secondary user on the primary user is calculated, and the secondary users whose comprehensive influence exceeds the comprehensive influence threshold are retained to obtain the initial social influence range of the primary user. Auxiliary users are screened according to their consumption tags within the initial social influence range of the primary user. If the consumption tags of the primary user and the auxiliary user are in the same category, or the upward price of the primary user is within the average price range of the auxiliary users, these auxiliary users are retained and other auxiliary users are eliminated to obtain the social influence range of the user.

3. The e-commerce user preference analysis system according to claim 2, characterized in that: The social module is used to: User behavior data also includes user consumption data, user browsing data, and user evaluation data; The interest tags of auxiliary users are determined through user social data, and the interest tags of auxiliary users are verified and adjusted based on user consumption data, user browsing data and user evaluation data, so as to determine the interest tags of each auxiliary user and build a social user interest list.

4. The e-commerce user preference analysis system according to claim 2, characterized in that: The system further comprises a comparison module, configured to: Before obtaining the user's personal interests under social influence according to the user's personal interests and the social user interest list, finding other user's personal interests similar to the user's personal interests, and adjusting the user's personal interests according to the other user's personal interests, including: Generate each user's personal interest tag based on user consumption data, user browsing data and user evaluation data, calculate the similarity between other users' personal interest tags and the main user's personal interest tags, and find other users with similar interests; If other users with similar interests are in the social network graph of the main user, the similar interests of the other users are regarded as explicit similar interests; Otherwise, the similar interests of the other users are regarded as invisible similar interests; Different adjustment weights are given to explicit similar interests and implicit similar interests to adjust the personal interest tags of the main user.

5. The e-commerce user preference analysis system according to claim 3, characterized in that: The interest module is used to: According to the user's personal interests and the social user interest list, the user's personal interests under social influence are obtained, and personal-social interest tags are generated Calculate the similarity between the adjusted personal interest tags of the main user and the interest tags in the social user's interest list, take the interest tags in the social user's interest list whose similarity exceeds the similarity threshold as similar social interest tags, and correct the personal interest tags of the main user by using the similar social interest tags, thereby generating personal-social interest tags; Among them, p i ′ is the value or weight of the i-th personal interest tag of the main user after similar social interest tags are corrected, p i is the value or weight of the i-th personal interest tag of the main user after adjustment of explicit similar interests and implicit similar interests, exp is the exponential function, n is the number of similar social interest tags whose similarity with the i-th personal interest tag of the main user exceeds the similarity threshold, α j is the social weight of the jth similar social interest tag similar to the i-th personal interest tag, C i，j is the similarity between the i-th personal interest tag and the j-th similar social interest tag, C 阈 is the similarity threshold, and k is a preset constant.

6. The e-commerce user preference analysis system according to claim 1, characterized in that: The interest module is used to: The first, second and third most popular indicators are selected through user behavior data, user social data and product data respectively; The first hot index, the second hot index and the third hot index are integrated and weighted to obtain a comprehensive hot index, and the hot product label and the unpopular product label of the product are generated according to the size of the comprehensive hot index.

7. The e-commerce user preference analysis system according to claim 1, characterized in that: The portrait module is used to: User consumption tags, personal-social interest tags, hot product tags and unpopular product tags are standardized, and weights are assigned to user consumption tags, personal-social interest tags, hot product tags and unpopular product tags. User consumption tags, personal-social interest tags, hot product tags and unpopular product tags are multi-label fused to build a multi-level user portrait.

Images