User preference analysis method and system

Abstract

The invention belongs to the technical field of data processing, and particularly relates to a user preference analysis method and system.The method comprises the steps that user interaction data of shops in a business district are obtained, and the user interaction data comprise interaction duration, collection, comment and other specific interaction behaviors; the interestingness is calculated by coding the interaction behaviors and comparing the interaction behaviors with standard interaction behaviors; weighting the original interaction duration by taking the interestingness as a weight so as to effectively distinguish effective interaction with high intention from invalid interaction with low intention; and calculating the comprehensive preference degree of the user by integrating the weighted interaction duration, the interaction times and the proportion of high-quality interaction. According to the invention, the accuracy and reliability of user preference analysis are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. More specifically, this invention relates to a user preference analysis method and system. Background Technology

[0002] In scenarios such as business operations, precision marketing, and urban planning, analyzing user preferences within specific business districts is a crucial step. By analyzing user interaction data with shops on online application platforms such as lifestyle service and map applications, it is possible to accurately depict user profiles, assess shop popularity, and provide decision-making basis for optimizing the business format and attracting investment to the business district.

[0003] In existing user preference analysis technologies, relatively direct indicators are usually used to measure the degree of user interest. For example, the duration or number of user interactions are used as the core indicators to measure user interest. The longer a user stays on a store page or the more times they visit it, the stronger their interest in the store and the higher their preference. By statistically analyzing and aggregating relevant indicators from a large number of users, stores in a business district can be ranked by popularity, thereby determining the overall trend of user preferences.

[0004] However, the accuracy of analysis results relying solely on interaction duration is insufficient. For example, users may accidentally enter a store page due to ad pushes, or they may not actually browse the page after opening it because they are handling other tasks, but the system will still record a long period of invalid interaction time. This kind of pseudo-high popularity data will interfere with the analysis results, and simply considering the number of interactions cannot distinguish the quality of interactions. For example, users who have engaged in in-depth interactions such as collecting and commenting usually have a higher degree of preference and decision intention than users who have only passively browsed multiple times. Therefore, the analysis of users' true preferences is not accurate enough. Summary of the Invention

[0005] To address the aforementioned technical problem of insufficient accuracy in user preference analysis, the present invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a user preference analysis method, comprising: The process involves: acquiring a sequence of shops within a target business district and their interaction data, including user interactions, interaction durations, and corresponding behaviors; encoding the interactions in each record and creating an interaction feature vector for each record; establishing a standard interaction feature vector; calculating the interest level of any user in any shop's interaction record based on the similarity between the user's interaction feature vector and the standard interaction feature vector; using the user's interest level as a weighting coefficient to weight the interaction duration of the user's interaction record with the shop, obtaining the weighted interaction duration of any user in any shop's interaction record; combining the number of interactions with a shop and the weighted interaction duration to obtain the user's overall interest level in any shop; combining the user's interest level with the proportion of interaction records where the user interacts with a shop to obtain the user's overall preference for the shop; and ranking the shops in the target business district based on each user's overall preference for each shop to obtain a shop heatmap, thus completing the analysis of user preferences within the target business district.

[0007] This invention encodes users' specific interactive behaviors such as collecting and commenting, and calculates the similarity between these behaviors and standard interactive behaviors to determine the interest level of each interaction. This interest level is then used as a weight to weight the interaction duration, effectively distinguishing between high-intent effective interactions and low-intent ineffective interactions. This invention integrates weighted interaction duration, interaction frequency, and the proportion of effective interaction records to construct a comprehensive preference level model, improving the accuracy and reliability of user preference analysis in business districts.

[0008] Preferably, encoding the interactive behavior of the interaction record and establishing the interaction feature vector of each interaction record includes: The user's interactive behavior is encoded. For the a-th type of interactive behavior, if the user has the a-th type of interactive behavior, the a-th type of interactive behavior is encoded as 1, and if the user does not have the a-th type of interactive behavior, the a-th type of interactive behavior is encoded as 0. All interactive behaviors of each user to each store's interactive records are encoded using 0 and 1, and the interactive feature vector of the corresponding interactive record is constructed according to the order of the interactive behaviors.

[0009] This invention transforms abstract user behaviors such as collecting, commenting, and forwarding into standardized digital feature vectors that can be directly processed by computers by using a binary method where 1 represents the presence of a user and 0 represents the absence of a user. This provides a foundation for subsequent calculation of interaction quality using mathematical methods, making the entire analysis process more standardized and easier to implement.

[0010] Preferably, the establishment of the standard interaction feature vector includes: Set a vector with the same order as the interaction feature vector and all feature values ​​of 1, and denote it as the standard interaction feature vector.

[0011] Preferably, the interest level of any user in any interaction record of any store satisfies the expression: ; In the formula, This represents the interest level of the i-th user in the k-th interaction record of the d-th shop; This represents the interaction feature vector of the k-th interaction record between the i-th user and the d-th shop; Represents the standard interactive feature vector; This represents the vector dot product operator; Represents the normalization function; This represents a tiny value and is used to avoid having a denominator of 0 in the expression.

[0012] This invention uses a normalized calculation method based on vector dot product, which not only considers the types of user interaction behaviors, but also reflects the degree of closeness between the combination pattern of interaction behaviors and the ideal pattern, thereby obtaining a more discriminative interest score and making the weight allocation more reasonable.

[0013] Preferably, the weighted interaction duration of any user's interaction record with any store satisfies the expression: ; In the formula, This represents the weighted interaction duration of the k-th interaction record between the i-th user and the d-th shop; This represents the duration of the k-th interaction record between the i-th user and the d-th shop; This represents the interest level of the i-th user in the k-th interaction record of the d-th shop; This represents the number of interactions between the i-th user and the d-th shop.

[0014] This invention normalizes the interest level of a single interaction by summing the interest levels of all interactions in the store to which it belongs, and then multiplies it by the original interaction duration to achieve weighted interaction duration. This allows the total interaction duration to be redistributed according to the interaction quality, amplifying the impact of the interaction duration corresponding to high interest interactions while weakening the impact of the interaction duration corresponding to low interest interactions. The resulting weighted interaction duration better reflects the user's true level of engagement.

[0015] Preferably, the interest level of any user in any shop satisfies the expression: ; In the formula, This represents the level of interest of the i-th user in the d-th shop; This represents the number of interactions between the i-th user and the d-th shop. This represents the set of interactions between the i-th user and all shops. This represents the average weighted interaction duration of all interaction records between the i-th user and the d-th shop; This represents the average weighted interaction time of the i-th user across all interaction records with all shops; This represents the maximum value function.

[0016] Preferably, obtaining the user's overall preference for the store includes: Calculate the degree of personal preference of the i-th user for the d-th shop; obtain the set of personal preference degrees of all users for all shops; The ratio of the degree of personal preference of the i-th user for the d-th shop to the maximum value of the set of personal preference degrees is denoted as the overall degree of personal preference of the i-th user for the d-th shop.

[0017] This invention solves the problem of incomparable preference scores caused by differences in activity levels and interaction habits among different users by comparing the individual preference level of a single user with the highest preference level among all users. It can eliminate the bias caused by individual differences and make the final comprehensive preference level a comparable indicator across the entire user group, thus providing a more objective basis for ranking the popularity of shops.

[0018] Preferably, the calculation of the degree of personal preference of the i-th user for the d-th shop includes: Acquire the number of high-quality interactions between each user and each store; Obtain the ratio of the number of high-quality interactions between the i-th user and the d-th shop to the number of interactions between the i-th user and the d-th shop; add the ratio to the i-th user's interest in the d-th shop and perform positive correlation normalization to obtain the i-th user's overall preference for the d-th shop.

[0019] Preferably, obtaining the number of high-quality interactions between each user and each store includes: The number of interaction records where the interaction feature vector is not zero is obtained from the interaction records of each user with each store. This number is recorded as the number of high-quality interactions of each user with the corresponding store.

[0020] Secondly, the present invention provides a user preference analysis system, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned user preference analysis method is implemented.

[0021] By adopting the above technical solution, a user preference analysis method is generated into a computer program and stored in a memory for loading and execution by a processor. This allows for the creation of a terminal device based on the memory and processor, facilitating its use.

[0022] The beneficial effects of this invention are as follows: by encoding users' specific interactive behaviors such as collections and comments into feature vectors, and calculating their similarity with standard interaction vectors to obtain the interest level of a single interaction, and using the interest level as a weight to weight the original interaction duration, invalid data such as accidental touches and idle time are effectively filtered out. Then, by comprehensively considering the weighted duration, number of interactions, and proportion of high-quality interactions, a more accurate and comprehensive user preference model is constructed, thereby improving the accuracy of user preference analysis in business districts. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a user preference analysis method according to the present invention; Figure 2 This is a schematic diagram illustrating the user's interaction history with the store; Figure 3 This is a schematic diagram illustrating the comparison between the interaction duration of a user's interaction record and the weighted interaction duration. Detailed Implementation

[0024] This invention discloses a user preference analysis method, referring to... Figure 1 This includes steps S1-S4: S1: Obtain the sequence of shops in the target business district and the interaction data of each shop. The interaction data includes several users with interaction records, the corresponding interaction duration, and the interaction behaviors.

[0025] It should be noted that in order to comprehensively and accurately analyze user preferences within a business district, it is necessary to analyze the users of all shops within the district in a timely and comprehensive manner. A business district big data management system can obtain specific shop information, while users access shop operating and service information through lifestyle application platforms, interact with these platforms, and then make decisions about the services they need. These interactions are crucial data for analyzing user preferences. Therefore, this invention acquires shop information within a business district and then obtains user interaction data from lifestyle application platforms containing these shops over a period of time, thereby providing an accurate data foundation for subsequent user preference analysis.

[0026] Specifically, the system obtains the shop sequence of the target business district through a business district big data management system. Then, it retrieves interaction data for all shops in the shop sequence within a preset time period from the databases of various lifestyle application platforms. For example, the preset time period is one month. It should be noted that user preference analysis needs to be timely. If the preset time period is too long, changes in shops may lead to biases in the user preference analysis; if the preset time period is too short, errors may occur in the user preference analysis of seasonally active shops. It should also be noted that the sorting method of the shop sequence is set by the implementers based on the actual implementation situation, for example, sorting by initial opening time.

[0027] The store's interaction data includes several users with interaction records, along with the corresponding interaction duration and the interaction behaviors. It should be noted that having interaction records indicates that the user has browsed the store; interaction behaviors include actions such as saving, forwarding, and commenting. Figure 2 This diagram illustrates user interactions with a store, including two types of interactions: favorites and comments. The horizontal axis represents the number of interactions, and the vertical axis represents the duration of the interactions.

[0028] At this point, the sequence of shops in the target business district and the interaction data of each shop have been obtained.

[0029] S2: Encode the interactive behavior of the interaction records and establish the interaction feature vector of each interaction record; establish a standard interaction feature vector; calculate the interest degree of any user on any interaction record of any store based on the similarity between the interaction feature vector of the user's interaction record with the standard interaction feature vector; use the user's interest degree on the store's interaction record as a weight coefficient to weight the interaction duration of the user's interaction record on the store, and obtain the weighted interaction duration of the user's interaction record on the store.

[0030] It should be noted that while interaction duration can directly measure user interest when analyzing user behavior—the longer a user stays on a store's page on the platform, the more attention and effort they invest, and the higher their level of interest—in real-world applications, relying solely on browsing time is insufficient. For example, users might accidentally click on ads or be unused on their phones for other reasons, yet the interaction duration is still being recorded, resulting in numerous invalid clicks and interactions. Therefore, it's necessary to analyze the validity of interactions within the user's interaction log. Considering that interactive behavior clearly indicates user interest, users will only bookmark stores and engage in similar activities when they are interested in them and have a specific purpose. Therefore, this invention combines the richness of user interaction behavior within the store's interaction log to calculate the user's level of interest in the store's interaction records.

[0031] It should be further explained that, within the same interaction duration, the more interactive behaviors occur, indicating higher user interest, and the more effective the corresponding interaction duration is for analyzing user preferences. For example... Figure 3 This diagram illustrates the comparison between the interaction duration of a user's interaction records and the weighted interaction duration, showing the difference between the weighted interaction duration after the interest-based weighting adjustment and the initial interaction duration. Therefore, this invention further uses the user's interest in the store's interaction records as a weighting criterion to weight the user's interaction duration, thereby improving the effectiveness of preference analysis for each interaction record and increasing the accuracy of the final preference analysis.

[0032] Specifically, the interactive behaviors of the interaction records are encoded, and an interaction feature vector is established for each interaction record. This includes encoding the user's interactive behaviors. For the 'a'-th type of interactive behavior, if the user exhibits the 'a'-th type of interactive behavior, then the 'a'-th type of interactive behavior is encoded as 1; otherwise, it is encoded as 0. Thus, 0 and 1 can be used to represent the interaction feature vector of a user's interaction record. It should be noted that, taking favorites, comments, and reposts as three interactive behaviors, if the 'i'-th user exhibits favorites and comments but not reposts in the 'k'-th interaction record of the 'd'-th shop, then the interaction feature vector of the 'i'-th user's 'k'-th interaction record of the 'd'-th shop is: .

[0033] Preferably, a standard interaction feature vector is established, including: setting a vector with the same order as the interaction feature vector and all eigenvalues ​​being 1, denoted as the standard interaction feature vector. It should be noted that, taking the three interaction behaviors of collecting, commenting, and forwarding as examples, the standard interaction feature vector is... This indicates that the user has engaged in all interactions with the store.

[0034] It should be noted that the closer the interaction feature vector of a user's interaction record with a store is to the standard interaction feature vector, the more interactive behaviors the user has exhibited with the store, indicating a higher level of user interest in the store. Therefore, based on the similarity between the interaction feature vector of a user's interaction record with a store and the standard interaction feature vector, the interest level of any user in any interaction record with any store can be calculated.

[0035] Preferably, the interest level of any user in any interaction record of any store satisfies the expression: ; In the formula, This represents the interest level of the i-th user in the k-th interaction record of the d-th shop; This represents the interaction feature vector of the k-th interaction record between the i-th user and the d-th shop; Represents the standard interactive feature vector; This represents the vector dot product operator; Represents the normalization function; To represent minute values, used to avoid zero denominators in expressions, for example when... hour.

[0036] In the formula, Let represent the dot product of the interaction feature vector of the i-th user's k-th browsing record of the d-th shop and the standard interaction feature vector. This represents the product of the magnitude of the interaction feature vector of the i-th user's k-th browsing record of the d-th shop and the magnitude of the standard interaction feature vector. express and The cosine of the angle between the two points represents the similarity between the interaction feature vector of the i-th user's k-th visit to the d-th shop and the standard interaction feature vector. The closer this value is to 1, the stronger the similarity. and The closer the cosine of the included angle is to 1, the better. and The closer the angle between the two points is to 0, the more similar the interaction feature vector of the i-th user's k-th visit to the d-th shop is to the standard interaction feature vector. For example, if the interaction feature vector is of order 3, it indicates that there are three types of interaction behaviors. The closer ,but The closer ; Yes Normalization is performed so that The value range is between 0 and 1.

[0037] Preferably, the user's interest in the store's interaction records is used as a weighting coefficient to weight the interaction duration of the user's interaction records with the store, obtaining the weighted interaction duration of the user's interaction records with the store, including: The weighted interaction duration of any user's interaction with any store at any time satisfies the expression: ; In the formula, This represents the weighted interaction duration of the k-th interaction record between the i-th user and the d-th shop; This represents the duration of the k-th interaction record between the i-th user and the d-th shop; This represents the interest level of the i-th user in the k-th interaction record of the d-th shop; This represents the number of interactions between the i-th user and the d-th shop.

[0038] In the formula, This represents the interest level of the i-th user in the k-th interaction record of the d-th shop, and is the proportion of interest of the i-th user in the total interaction with the d-th shop. The larger this value, the higher the proportion of interest of the i-th user in the k-th interaction record of the d-th shop, and the greater the weighting of the k-th interaction record of the i-th user in the d-th shop on the weighted interaction duration. This means that the interaction duration of the k-th interaction record of the i-th user with the d-th shop is numerically redistributed using the normalized interest score of the i-th user with the d-th shop as the weight. When the interaction duration of the i-th user with the k-th interaction record of the d-th shop is long, but the corresponding interest score is low, it means that the effectiveness of the interaction during this period is not high. Therefore, the weighted interaction duration of the i-th user with the k-th interaction record of the d-th shop is reduced to balance the referenceability of different interaction records.

[0039] At this point, the weighted interaction duration of the user's interaction records with the store has been obtained.

[0040] S3: Combine the number of user interactions with a store and the weighted interaction duration to obtain the user's interest in any store; combine the user's interest in a store with the proportion of user interaction records that show interaction behavior with the store to obtain the user's overall preference for the store.

[0041] It should be noted that for an individual user, among all the shops they have interacted with, the user shows a higher level of interest in shops with a longer weighted interaction time. However, if the user only has one corresponding interaction record, it indicates that the interest is short-lived. Therefore, the user's interest in each shop should also be obtained by comparing the number of interactions with all shops.

[0042] It should be further explained that, considering the actual number of times users browse each store may vary—for example, user B might visit a store 19 times and add it to their favorites once, investing more time—while user A might visit the store 3 times and add it to their favorites once, each visit being more valuable, resulting in a higher efficiency in converting user decision-making intentions than user B. Therefore, this invention incorporates the effective interaction rate of user interactions with stores to comprehensively obtain the degree of user preference for a store.

[0043] Specifically, by combining the number of user interactions with a store and the weighted interaction duration, the interest level of any user in any store can be obtained, including: It should be noted that in any user's interaction history, the higher the relative number of interactions and the longer the relative weighted interaction duration for any given store, the greater the user's interest in that store. Therefore, by combining the number of interactions and the weighted interaction duration, the interest level of any user in any given store can be obtained.

[0044] The expression that any user's interest in any shop satisfies: ; In the formula, This represents the level of interest of the i-th user in the d-th shop; This represents the number of interactions between the i-th user and the d-th shop. This represents the set of interactions between the i-th user and all shops. This represents the average weighted interaction duration of all interaction records between the i-th user and the d-th shop; This represents the average weighted interaction time of the i-th user across all interaction records with all shops; This represents the maximum value function.

[0045] In the formula, This represents the ratio of the number of interactions between the i-th user and the d-th shop to the maximum number of interactions between all shops. This value reflects the concentration of interactions between the i-th user and the d-th shop. The larger the value, the closer the i-th user's interactions with the d-th shop are to their maximum interaction level. This means that the d-th shop has a sustained attraction for the i-th user, prompting them to visit and view it repeatedly, reflecting the breadth of the user's interests. This represents the ratio of the average weighted interaction time of the i-th user with the d-th shop to the average weighted interaction time of the i-th user with all shops. This value reflects the i-th user's browsing focus on the d-th shop. The larger the value, the more time and effort the i-th user invests each time they visit the d-th shop, and the closer it is to their highest level of focus. This means that the content and characteristics of the shop are highly attractive to the user. This means that the attraction of the i-th user to the d-th shop is combined with the browsing focus. When both the attraction and browsing focus of the i-th user to the d-th shop are large, it reflects the high level of interest of the i-th user in the d-th shop.

[0046] It should be noted that the higher the proportion of user interaction records, the higher the user's preference for the corresponding store. Therefore, by combining the user's interest in the store with the proportion of user interaction records related to the store, the user's preference for the store can be obtained.

[0047] Preferably, by combining user interest in a store with the proportion of user interaction records showing interaction with the store, the overall user preference for the store is obtained, including: The number of interaction records where the interaction feature vector is not zero is obtained from the interaction records of each user with each store. This number is recorded as the number of high-quality interactions of each user with the corresponding store.

[0048] The overall preference of any user for any shop satisfies the expression: ; ; In the formula, This represents the overall preference level of the i-th user for the d-th shop; This represents the degree of personal preference of the i-th user for the d-th shop; This represents the set of all users' personal preferences for all shops; This represents the number of high-quality interactions between the i-th user and the d-th shop. This represents the number of interactions between the i-th user and the d-th shop. This represents the level of interest of the i-th user in the d-th shop; This represents the maximum value function.

[0049] In the formula, This represents the percentage of high-quality interactions between user i and store d out of all interactions. A larger value indicates a deeper and more effective interaction between user i and store d. Based on this, combined with user i's interest in store d, it indicates that when user i has a deep interaction with store d and a high level of interest, user i has a greater preference for store d.

[0050] In the formula, This represents the ratio of the degree of personal preference of the i-th user for the d-th shop to the maximum value of the degree of personal preference of all users for all shops. This value means that the preference of the i-th user for the d-th shop is included among all users. This avoids the bias in the overall user preference analysis of shops caused by the differences in activity levels, behavioral habits and other factors among different users due to the analysis of only individuals, which would lead to errors in the final user preference analysis conclusion.

[0051] At this point, we have obtained the overall preference level of each user for each store.

[0052] S4: Based on each user's overall preference for each shop, sort the shops in the target business district, obtain the shop heat sequence, and complete the analysis of user preferences within the target business district.

[0053] It should be noted that by periodically obtaining the overall preferences of users of all shops in the target business district, we can analyze changes in user preferences and provide a basis for the site selection of other related shops and chain stores.

[0054] Specifically, by accumulating the overall preference levels of each shop, we can obtain the overall popularity of each shop, form a shop popularity sequence of the business district, and then promote the higher-ranked shops with a larger proportion, and designate vacant shops as shops of the relevant type, thereby completing the analysis of user preferences in the business district.

[0055] It should be noted that the analysis of user preferences in a business district can also analyze the preferences of users in different shops. For example, users in nearby office buildings may have preferences for daily dining. The overall preference of users for shops provides a basic and accurate data foundation for the analysis of user preferences in a business district.

[0056] This concludes the user preference analysis for the target business district.

[0057] This invention also discloses a user preference analysis system, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a user preference analysis method according to the present invention.

[0058] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0059] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A user preference analysis method, characterized in that, include: Obtain the sequence of shops in the target business district and the interaction data of each shop. The interaction data includes several users with interaction records, the corresponding interaction duration, and the interaction behaviors. The interactive behaviors of the interaction records are encoded, and an interaction feature vector for each interaction record is established. Establish a standard interaction feature vector; calculate the interest of any user in any interaction record of any store based on the similarity between the interaction feature vector of the user's interaction record with the standard interaction feature vector; use the interest of the user in the interaction record of the store as a weight coefficient to weight the interaction duration of the user's interaction record of the store, and obtain the weighted interaction duration of any user in any interaction record of any store. By combining the number of user interactions with a store and the weighted interaction duration, we can obtain the interest level of any user for any store. By combining users' interest in shops with the proportion of user interaction records that show interaction behavior with shops, we can obtain users' overall preference for shops. Based on each user's overall preference for each shop, the shops in the target business district are ranked to obtain a shop heat map, thus completing the analysis of user preferences within the target business district.

2. The user preference analysis method according to claim 1, characterized in that, The process of encoding the interactive behavior of the interaction records and establishing the interaction feature vector for each interaction record includes: The user's interactive behavior is encoded. For the a-th type of interactive behavior, if the user has the a-th type of interactive behavior, the a-th type of interactive behavior is encoded as 1, and if the user does not have the a-th type of interactive behavior, the a-th type of interactive behavior is encoded as 0. All interactive behaviors of each user to each store's interactive records are encoded using 0 and 1, and the interactive feature vector of the corresponding interactive record is constructed according to the order of the interactive behaviors.

3. The user preference analysis method according to claim 1, characterized in that, The establishment of the standard interactive feature vector includes: Set a vector with the same order as the interaction feature vector and all feature values ​​of 1, and denote it as the standard interaction feature vector.

4. The user preference analysis method according to claim 1, characterized in that, The interest level of any user in any interaction record with any store satisfies the expression: ； In the formula, This represents the interest level of the i-th user in the k-th interaction record of the d-th shop; This represents the interaction feature vector of the k-th interaction record between the i-th user and the d-th shop; Represents the standard interactive feature vector; This represents the vector dot product operator; Represents the normalization function; This represents a tiny value and is used to avoid having a denominator of 0 in the expression.

5. The user preference analysis method according to claim 1, characterized in that, The weighted interaction duration of any user's interaction record with any store satisfies the expression: ； In the formula, This represents the weighted interaction duration of the k-th interaction record between the i-th user and the d-th shop; This represents the duration of the k-th interaction record between the i-th user and the d-th shop; This represents the interest level of the i-th user in the k-th interaction record of the d-th shop; This represents the number of interactions between the i-th user and the d-th shop.

6. The user preference analysis method according to claim 1, characterized in that, The interest level of any user in any shop satisfies the expression: ； In the formula, This represents the level of interest of the i-th user in the d-th shop; This represents the number of interactions between the i-th user and the d-th shop. This represents the set of interactions between the i-th user and all shops. This represents the average weighted interaction duration of all interaction records between the i-th user and the d-th shop; This represents the average weighted interaction time of the i-th user across all interaction records with all shops; This represents the maximum value function.

7. The user preference analysis method according to claim 1, characterized in that, The process of obtaining users' overall preference for shops includes: Calculate the degree of personal preference of the i-th user for the d-th shop; obtain the set of personal preference degrees of all users for all shops; The ratio of the degree of personal preference of the i-th user for the d-th shop to the maximum value of the set of personal preference degrees is denoted as the overall degree of personal preference of the i-th user for the d-th shop.

8. The user preference analysis method according to claim 7, characterized in that, The calculation of the degree of personal preference of the i-th user for the d-th shop includes: Acquire the number of high-quality interactions between each user and each store; Obtain the ratio of the number of high-quality interactions between the i-th user and the d-th shop to the number of interactions between the i-th user and the d-th shop; add the ratio to the i-th user's interest in the d-th shop and perform positive correlation normalization to obtain the i-th user's overall preference for the d-th shop.

9. The user preference analysis method according to claim 8, characterized in that, The method of obtaining the number of high-quality interactions between each user and each store includes: The number of interaction records where the interaction feature vector is not zero is obtained from the interaction records of each user with each store. This number is recorded as the number of high-quality interactions of each user with the corresponding store.

10. A user preference analysis system, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a user preference analysis method according to any one of claims 1-9.

Images