A method and system for user preference analysis

Abstract

This application relates to the technical field of big data processing, and in particular to a user preference analysis method and system. The method includes: inputting the heat value sequences of the target user's various requirements into a long short-term memory network to obtain the long-term features and short-term features of each requirement, and the long-term features of each requirement form the user portrait of the target user; obtaining the correlation matrix of the user group to which the target user belongs, and inputting the short-term features of each requirement and the correlation matrix into a preference analysis model to obtain the heat attenuation coefficient of each requirement; inputting the real-time heat value and heat attenuation coefficient of each requirement of the target user into a preset attenuation model to predict the heat duration of each requirement, and obtaining the analysis result. Through the technical solution of this application, it is possible to take into account the heat attenuation of the user for each requirement and obtain accurate preference analysis results.

Description

Technical Field

[0001] This application relates to the technical field of big data processing, and particularly relates to a user preference analysis method and system. Background Art

[0002] With the advent of the big data era, a variety of services such as goods, financial services, and multimedia content can be provided on the Internet. In order to recommend services of interest to users, it is necessary to analyze user preferences to improve the user experience when using the Internet.

[0003] Currently, the patent application document with the publication number CN114840486A discloses a user behavior data collection method, system, and cloud platform. The method includes: parsing the user behavior data set in the Internet financial business session log through a user behavior capture strategy, separating the behavior event session log corresponding to the user behavior data set, and then mining the event interest points of the behavior event session log through an interest point mining strategy to obtain user interest knowledge fields. It can accurately and efficiently parse the user behavior data set from Internet financial business session logs in various forms and collection methods, and can also specifically parse the user interest knowledge fields from the behavior event session logs corresponding to the user behavior data set, and can achieve accurate classification of the interest knowledge categories of a large number of user behavior events.

[0004] The above method mines the event interest points of the behavior event session log to obtain the user interest knowledge fields, thereby judging the knowledge categories that the user is interested in. However, the user's interest level in knowledge categories changes gradually. Ignoring the change in the interest level during the analysis of user preferences will lead to inaccurate results of user preference analysis. Summary of the Invention

[0005] To solve the technical problem of inaccurate user preference analysis results, this application provides a user preference analysis method and system, which can take into account the heat attenuation of each user demand and obtain accurate preference analysis results.

[0006] In the first aspect of the present application, a user preference analysis method is provided. The analysis method includes: inputting the heat value sequences of each requirement of a target user into a long short-term memory network to obtain the long-term features and short-term features of each requirement, and the long-term features of each requirement constitute the user profile of the target user; clustering the user profiles of all users to obtain the user group to which the target user belongs; obtaining the correlation matrix of the user group, and inputting the short-term features of each requirement and the correlation matrix into a preference analysis model to obtain the heat attenuation coefficient of each requirement, where the correlation matrix includes the correlation of the change in heat values between any two requirements; inputting the real-time heat value and the heat attenuation coefficient of each requirement of the target user into a preset attenuation model to predict the heat duration of each requirement, and taking the heat duration of each requirement as the analysis result; the training method of the preference analysis model includes: obtaining the short-term features of each requirement and the correlation matrix of the user group to which the user belongs according to the first historical sequence of the heat values of each requirement of any user as a set of training samples; fitting a preset attenuation model in the second historical sequence of the heat values of each requirement after the first historical sequence of the user heat values to obtain the attenuation coefficient label of the training samples; inputting the training samples into the preference analysis model to obtain an output result, and calculating the mean square error loss function of the output result and the attenuation coefficient label to iteratively train the preference analysis model.

[0007] First, input the heat value sequences of each requirement of the target user into a long short-term memory network to obtain the long-term features and short-term features of each requirement, and use the long-term features of each requirement as the user profile of the target user. This user profile can reflect the long-term trend of the heat values of all requirements of the target user; cluster the user profiles of all users to obtain the user group to which the target user belongs, obtain the correlation matrix of the user group, and calculate the correlation of the change in heat values between any two requirements within the user group, which can avoid the correlation differences brought by different user groups and improve the accuracy of the correlation matrix; input the short-term features of each requirement and the correlation matrix into a preference analysis model to obtain the heat attenuation coefficient of each requirement; input the real-time heat value and the heat attenuation coefficient of each requirement of the target user into a preset attenuation model to predict the heat duration of each requirement, and take the heat duration of each requirement as the analysis result. This analysis result takes into account the heat attenuation of the user for each requirement and improves the accuracy of the preference analysis result.

[0008] Preferably, the method for obtaining the heat value includes: obtaining various behavior data of the target user for any requirement within a collection period, where the behavior data includes the number of clicks, browsing duration, and number of purchases; determining the behavior scores of each behavior data according to a preset score interval, and weighted summing the behavior scores according to a preset weight to obtain the heat value of the target user for the requirement within the collection period.

[0009] Preferably, cluster the user portraits of all users to obtain the user group to which the target user belongs, including: clustering the user portraits using the Kmeans algorithm, determining the number of clustering clusters using the elbow method to obtain multiple clustering clusters and the clustering centers of each clustering cluster; calculating the Euclidean distance between the user portrait of the target user and each clustering center, and taking the clustering cluster to which the clustering center corresponding to the minimum Euclidean distance belongs as the user group to which the target user belongs.

[0010] Cluster the user portraits of all users, so that the multiple user portraits within a clustering cluster have small differences, and further determine the user group to which the target user belongs. The long-term trend of the heat value of each user in this user group is basically the same as that of the target user. Subsequently, calculating the correlation between the changes in heat values between any demands within the user group can avoid the correlation differences brought by different user groups and improve the accuracy of the correlation matrix.

[0011] Preferably, demand and demand The calculation method of the correlation between the changes in heat values includes: for any user in the user group, calculate the difference in heat values between the current collection period and the previous adjacent collection period to obtain the heat value change sequence of each demand, and calculate the Pearson correlation coefficient of the heat value change sequence between the user in demand and demand ; calculate the representativeness of the user, and the representativeness is negatively correlated with the Euclidean distance between the user portrait and the clustering center of the user group; perform weighted summation of the Pearson correlation coefficients of all users according to the representativeness to obtain the correlation between the changes in heat values between demand and demand .

[0012] The correlation matrix can reflect the correlation between the changes in heat values between any demands, can provide auxiliary information for predicting the heat decay coefficient of each demand, and taking the correlation matrix as the input of the preference analysis model can improve the accuracy of the heat decay coefficient.

[0013] Preferably, the representativeness of user is: , is the user portrait of user in the user group, is the clustering center of the user group, is and 's Euclidean distance, is the sum of the Euclidean distances between the user portraits of the users in the user group and .

[0014] There are also differences among multiple user portraits within a user group. The representativeness of each user within the user group is calculated based on the Euclidean distance between the user portrait and the clustering center of the user group. The greater the representativeness of a user, the greater the impact on relevance, improving the accuracy of relevance.

[0015] Preferably, the preference analysis model includes a convolutional sub-model and a regression sub-model; the convolutional sub-model is used to extract features from the correlation matrix to obtain correlation features; after splicing the correlation features and the short-term features of each requirement, they are input into the regression sub-model to output the heat decay coefficient of each requirement.

[0016] Preferably, predicting the heat duration of each requirement includes: inputting the real-time heat value and the heat decay coefficient of any requirement of the target user into a preset decay model to obtain the predicted heat values of the requirement in multiple future collection cycles; taking the first future collection cycle with a predicted heat value less than the heat threshold as the cut-off cycle, and taking the number of collection cycles between the current collection cycle and the cut-off cycle as the heat duration of the requirement.

[0017] The preset decay model is: , is the real-time heat value, is the heat decay coefficient, is the th collection cycle, is the th collection cycle's predicted heat value.

[0018] The calculation method of the mean square error loss function includes: calculating the average variance of the heat values of each requirement in the user group to which the user belongs; the mean square error loss function satisfies:

[0019] , is the number of all requirements, is the heat decay coefficient of requirement in the decay coefficient label, is the heat decay coefficient of requirement in the output result, is the heat value average variance of requirement , the sum of the heat value average variances of all requirements, is the sum of the absolute values of the heat decay coefficients in the output result.

[0020] In the user group to which the user belongs, if the long-term trend of requirement is stable and unchanged, it indicates that the user group can maintain a long-term interest in requirement , that is, the user has a long-term interest in requirement The heat attenuation coefficient should tend to 0. Conversely, if the demand has a large long-term fluctuation trend, it indicates that this user group cannot maintain a long-term interest in the demand , that is, the user's heat attenuation coefficient for the demand should be a large value. Therefore, the stability of the heat values of each demand in the user group to which the user belongs can reflect the relative magnitudes of the heat attenuation coefficients of each demand, providing auxiliary information for the training of the preference analysis model and ensuring the accuracy of the output results of the preference analysis model.

[0021] In the second aspect of the present application, a user preference analysis system is further provided, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a user preference analysis method according to the first aspect of the present application is implemented.

[0022] The technical solution of the present application has the following beneficial technical effects:

[0023] First, the heat value sequences of each demand of the target user are input into the long short-term memory network to obtain the long-term features and short-term features of each demand, and the long-term features of each demand are used as the user portrait of the target user. This user portrait can reflect the long-term trend of the heat values of all demands of the target user; cluster the user portraits of all users to obtain the user group to which the target user belongs, obtain the correlation matrix of the user group, and calculate the correlation of the change in heat values between any demands within the user group, which can avoid the correlation differences brought by different user groups and improve the accuracy of the correlation matrix; input the short-term features of each demand and the correlation matrix into the preference analysis model to obtain the heat attenuation coefficient of each demand; input the real-time heat value and heat attenuation coefficient of each demand of the target user into the preset attenuation model to predict the heat duration of each demand, and use the heat duration of each demand as the analysis result. This analysis result takes into account the heat attenuation of each demand by the user and improves the accuracy of the preference analysis result. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] By referring to the accompanying drawings and reading the following detailed description, the above and other objects, features, and advantages of the exemplary embodiments of the present application will become easy to understand. In the drawings, several embodiments of the present application are shown in an exemplary rather than restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts, where:

[0025] Figure 1 is a flowchart of a user preference analysis method according to an embodiment of the present application;

[0026] Figure 2 is a structural diagram of a preference analysis model according to an embodiment of the present application;

[0027] Figure 3 is a flowchart of a training method for a preference analysis model according to an embodiment of the present application;

[0028] Figure 4 is a structural block diagram of a user preference analysis system according to an embodiment of the present application. Detailed implementation manners

[0029] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments in the present application without creative efforts shall fall within the protection scope of the present application.

[0030] It should be understood that when terms such as "first" and "second" are used in the claims, the description, and the drawings of the present application, they are only used to distinguish different objects and not to describe a specific order. The terms "including" and "comprising" used in the description and claims of the present application indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0031] According to a first aspect of the present application, the present application provides a user preference analysis method for performing preference analysis on all users within an Internet platform, where the Internet platform can be a commodity sales platform or a financial service platform.

[0032] Figure 1 is a flowchart of a user preference analysis method according to an embodiment of the present application. As Figure 1 shown, the user preference analysis method includes steps S101 to S104, which are described in detail below.

[0033] S101: Input the heat value sequence of each requirement of the target user into a long short-term memory network to obtain the long-term feature and short-term feature of each requirement, and the long-term features of each requirement form the user portrait of the target user.

[0034] In one embodiment, the target user is any registered user of the Internet platform, and the method for obtaining the heat value includes: obtaining various behavior data of the target user for any requirement within a collection period, where the behavior data includes the number of clicks, browsing duration, and the number of purchases; determining the behavior scores of each behavior data according to a preset score interval, and weighted summing the behavior scores according to a preset weight to obtain the heat value of the target user for the requirement within the collection period.

[0035] Among them, the collection period can be 1 day, that is, the heat value of a certain demand of the target user can be collected every day. By arranging all the heat values in the order of the collection period, the heat value sequence of the target user for this demand can be obtained. The demand can be a commodity, a financial product or a multimedia video, and this application does not make any restrictions.

[0036] The preset score intervals for each behavior data are preset in advance; taking the number of clicks as an example, the preset score interval for a behavior score of 0 is ; the preset score interval for a behavior score of 1 is ; the preset score interval for a behavior score of 2 is ; the preset score interval for a behavior score of 3 is . Therefore, when the number of clicks is 7, the behavior score of the number of clicks is 1.

[0037] After obtaining the behavior scores of each behavior data, weighted sum the behavior scores according to the preset weights. The preset weights of the number of clicks, the browsing duration and the number of purchases are , and respectively; thus, the heat value of the target user for any demand within one collection period can be obtained.

[0038] In one embodiment, the collection period is set to 1 day. The heat value sequence of any demand of the target user includes the heat values of each day within the time period from the registration date to the current day. Among them, the heat value of the current day is the real-time heat value; input the heat value sequence of any demand of the target user into the long short-term memory network to obtain the long-term feature and the short-term feature of this demand. Among them, the long-term feature is used to characterize the long-term trend of the heat value of this demand of the target user since the registration date, and the short-term feature is used to characterize the short-term trend of the heat value of this demand of the target user. Since the long-term feature can characterize the long-term trend of the heat value of each demand of the target user, therefore, by splicing the long-term features of each demand together, the user portrait of the target user can be obtained.

[0039] It can be understood that both the long-term feature and the short-term feature are vectors with 1 row and A columns. The value of A is related to the long short-term memory network. Here, the long short-term memory network can adopt a pre-trained long short-term memory network for feature extraction of the heat value sequence. If the number of all demand types is B, then the user portrait of the target user is a matrix with A rows and B columns.

[0040] In this way, use the long short-term memory network to extract features from the heat value sequence of each demand of the target user, and construct the user portrait of the target user according to the long-term features of each demand extracted. This user portrait can reflect the long-term trend of the heat value of all demands of the target user.

[0041] S102. Cluster the user portraits of all users to obtain the user group to which the target user belongs.

[0042] In one embodiment, due to the large number of users on the Internet platform, each user corresponds to a user portrait. By clustering all user portraits, the user group to which the target user belongs can be determined, and the long-term trends of the heat values of each user within the user group are basically the same.

[0043] Specifically, clustering the user portraits of all users to obtain the user group to which the target user belongs includes: clustering the user portraits using the Kmeans algorithm, determining the number of clustering clusters using the elbow method to obtain multiple clustering clusters and the clustering centers of each clustering cluster; calculating the Euclidean distance between the user portrait of the target user and each clustering center, and taking the clustering cluster to which the clustering center corresponding to the minimum Euclidean distance belongs as the user group to which the target user belongs.

[0044] Among them, the elbow method is a well-known technique for those skilled in the art and will not be elaborated here. A clustering cluster includes at least one user portrait, and the clustering center is the average value of all user portraits in the belonging clustering cluster.

[0045] S103. Obtain the correlation matrix of the user group, and input the short-term features of each demand and the correlation matrix into the preference analysis model to obtain the heat decay coefficient of each demand. The correlation matrix includes the correlation of the heat value changes between any demands.

[0046] In one embodiment, multiple users within the user group have the same interest preferences. Calculating the correlation of the heat value changes between any demands within the user group can avoid the correlation differences brought by different user groups and improve the accuracy of the correlation matrix.

[0047] Specifically, the calculation method of the correlation of the heat value changes between demand and demand includes: for any user in the user group, calculate the difference in the heat value between the current collection period and the previous adjacent collection period to obtain the heat value change sequence of each demand, and calculate the Pearson correlation coefficient of the heat value change sequence of the user between demand and demand ; calculate the representativeness of the user, and the representativeness is negatively correlated with the Euclidean distance between the user portrait and the user group clustering center; perform a weighted sum of the Pearson correlation coefficients of all users according to the representativeness to obtain the correlation of the heat value changes between demand and demand .

[0048] Among them, the representativeness of user is: , For a user in the user group user profile is the clustering center of the user group is and Euclidean distance is the sum of the Euclidean distances between the user profile of the user in the user group and

[0049] Among them, the requirements and the requirement satisfy:

[0050] , is the number of users in the user group is the representativeness of the user is the user among the requirements and the requirement Pearson correlation coefficient of the heat value change sequence

[0051] The correlation matrix can reflect the correlation of the heat value changes between any requirements, can provide auxiliary information for predicting the heat decay coefficient of each requirement, and taking the correlation matrix as the input of the preference analysis model can improve the accuracy of the heat decay coefficient

[0052] In one embodiment, the preference analysis model is the short-term features and correlation matrix of each requirement, and the output is the heat decay coefficient of each requirement Figure 2 is the structural diagram of the preference analysis model according to the embodiment of the present application. The preference analysis model includes a convolutional sub-model and a regression sub-model; the convolutional sub-model is used to extract features from the correlation matrix to obtain correlation features; after splicing the correlation features and the short-term features of each requirement, input them into the regression sub-model, and output the heat decay coefficient of each requirement

[0053] Among them, the convolutional sub-model can adopt existing convolutional neural networks such as ResNet or VGGNet, and the regression sub-model can adopt a fully connected neural network

[0054] S104, input the real-time heat value and heat decay coefficient of each requirement of the target user into a preset decay model to predict the heat duration of each requirement, and take the heat duration of each requirement as the analysis result

[0055] In one embodiment, after obtaining the heat decay coefficient of each requirement by using the preference analysis model, according to the real-time heat value of each requirement of the target user and the preset decay model, the heat duration of the target user in each requirement can be predicted

[0056] ​​Specifically, predicting the duration of the popularity of each demand includes: inputting the real-time popularity value and the popularity decay coefficient of any demand of the target user into a preset decay model to obtain the predicted popularity values of the demand in multiple future collection cycles; taking the first future collection cycle with a predicted popularity value less than the popularity threshold as the cut-off cycle, and taking the number of collection cycles between the current collection cycle and the cut-off cycle as the duration of the popularity of the demand.

[0057] Among them, the preset decay model is: , is the real-time popularity value, is the popularity decay coefficient, is the th collection cycle, is the th predicted popularity value of the collection cycle.

[0058] The value of the popularity threshold is 1.5. After inputting the real-time popularity value and the popularity decay coefficient of any demand into the preset decay model, the popularity decay curve of the demand can be determined, and the predicted popularity values of the demand in multiple future collection cycles can be obtained. Taking the first future collection cycle with a predicted popularity value less than the popularity threshold as the cut-off cycle, the duration of the popularity of the demand of the target user can be obtained.

[0059] Taking the duration of the popularity of each demand as the analysis result, if the duration of the popularity of a demand is longer, it means that the target user will maintain a long-term interest in the demand, then the recommendation degree of the demand is greater, and the interested demands can be pushed to the target user according to the analysis result.

[0060] It should be noted that the average value of all predicted popularity values between the current collection cycle and the cut-off cycle of any demand can also be calculated, and the product of the average value and the duration of the popularity is used as the analysis result.

[0061] In one embodiment, in order to enable the preference analysis model to output an accurate popularity decay coefficient, the preference analysis model needs to be trained. Please refer to Figure 3 , which is a flowchart of the training method of the preference analysis model according to the embodiment of the present application. As Figure 3 shown, the user preference analysis method includes steps S201 to S203, which are described in detail below.

[0062] S201, obtaining the short-term features of each demand and the correlation matrix of the user group to which the user belongs according to the first historical sequence of the popularity values of each demand of any user as a set of training samples.

[0063] Among them, the method for obtaining the short-term features of each requirement and the correlation matrix of the user group to which the user belongs has been described in detail in steps S101 to S103, and will not be elaborated here.

[0064] S202. Fit a preset decay model to the second historical sequence of the heat value of each requirement after the first historical sequence of the user heat value, and obtain the decay coefficient label of the training sample.

[0065] Among them, for any requirement, take the last heat value in the first historical sequence of heat values as the real-time heat value in the preset decay model; fit the preset decay model to the second historical sequence of heat values to determine the heat decay coefficient of the requirement; the heat decay coefficients of all requirements constitute the decay coefficient label of the training sample.

[0066] Exemplarily, the first historical sequence of heat values is the sequence of heat values of each requirement collected during the 1st collection period to the 15th collection period; the second historical sequence of heat values is the sequence of heat values of each requirement collected during the 16th collection period to the 25th collection period after the first historical sequence of heat values; for requirement , take the heat value of the 15th collection period as the real-time heat value , and use the least squares method to fit the preset decay model to the second historical sequence of heat values, then the heat decay coefficient of requirement can be determined.

[0067] S203. Input the training sample into the preference analysis model to obtain the output result, and calculate the mean square error loss function of the output result and the decay coefficient label to iteratively train the preference analysis model.

[0068] Among them, using the gradient descent method to iteratively train the preference analysis model is a well-known technique to those skilled in the art and will not be elaborated here. The preference analysis model is iteratively trained multiple times until the value of the mean square error loss function is less than the preset loss, and the training of the preference analysis model is completed. Among them, the preset loss value is 0.01.

[0069] The mean square error loss function satisfies: , is the number of all requirements, is the heat decay coefficient of requirement in the decay coefficient label, is the heat decay coefficient of requirement in the output result.

[0070] It should be noted that since different user groups have different long-term trends in the heat values of each requirement. Among the user groups to which the user belongs, if requirement If the long-term trend is stable and unchanged, it indicates that this user group can maintain a long-term interest in the demand That is to say, the heat decay coefficient of the demand for the user should tend to 0. On the contrary, if the long-term trend of the demand is relatively volatile, it indicates that this user group cannot maintain a long-term interest in the demand That is, the heat decay coefficient of the demand for the user should be a relatively large value. Therefore, the stability of the heat value of each demand in the user group to which the user belongs can be calculated. This heat value stability can reflect the relative magnitudes among the heat decay coefficients of each demand, providing auxiliary information for the training of the preference analysis model and ensuring the accuracy of the output result of the preference analysis model.

[0071] Specifically, the calculation method of the mean square error loss function includes: calculating the average variance of the heat values of each demand in the user group to which the user belongs; the mean square error loss function satisfies:

[0072] , is the number of all demands, is the heat decay coefficient of demand in the decay coefficient label, is the heat decay coefficient of demand in the output result, is the average variance of the heat value of demand , is the sum of the average variances of the heat values of all demands, is the sum of the absolute values of the heat decay coefficients in the output result.

[0073] Among them, the calculation process of the average variance of the heat value of demand is as follows: in the user group to which the user belongs, calculate the variance of the heat value sequence of each user demand , and take the average of all variances as the average variance of the heat value of demand .

[0074] The technical principle and implementation details of a user preference analysis method according to the present application are introduced through specific embodiments. First, the heat value sequences of each requirement of the target user are input into the long short-term memory network to obtain the long-term features and short-term features of each requirement, and the long-term features of each requirement are used as the user portrait of the target user, which can reflect the long-term trend of the heat values of all requirements of the target user; clustering the user portraits of all users to obtain the user group to which the target user belongs, obtaining the correlation matrix of the user group, and calculating the correlation between the changes in heat values between any two requirements within the user group can avoid the correlation differences brought by different user groups and improve the accuracy of the correlation matrix; inputting the short-term features of each requirement and the correlation matrix into the preference analysis model to obtain the heat attenuation coefficient of each requirement; inputting the real-time heat value of each requirement of the target user and the heat attenuation coefficient into a preset attenuation model to predict the heat duration of each requirement, and using the heat duration of each requirement as the analysis result, which takes into account the heat attenuation of the user for each requirement and improves the accuracy of the preference analysis result.

[0075] According to the second aspect of the present application, the present application further provides a user preference analysis system. Figure 4 It is a structural block diagram of a user preference analysis system according to an embodiment of the present application. As Figure 4 shown, the system 50 includes a processor and a memory, and the memory stores computer program instructions, which, when executed by the processor, implement a user preference analysis method according to the first aspect of the present application. The system also includes other components well known to those skilled in the art such as a communication bus and a communication interface, and their settings and functions are known in the art, so they will not be elaborated here.

[0076] In this application, the aforementioned memory can be any tangible medium that contains or stores a program, which can be used by or in conjunction with an instruction execution system, system, or device. For example, a computer-readable storage medium can be any suitable magnetic storage medium or magneto-optical storage medium, such as resistive random access memory (RRAM), dynamic random access memory (DRAM), static random-access memory (SRAM), enhanced dynamic random access memory (EDRAM), high-bandwidth memory (HBM), hybrid memory cube (HMC), etc., or any other medium that can be used to store the required information and can be accessed by an application, module, or both. Any such computer storage medium can be part of the device or accessible or connectable to the device. Any application or module described in this application can be implemented by computer-readable / executable instructions stored or otherwise maintained by such a computer-readable medium.

[0077] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0078] The above-described embodiments merely represent several implementation manners of this application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent application. It should be noted that for those of ordinary skill in the art, without departing from the concept of this application, several modifications and improvements can still be made, and these all belong to the protection scope of this application. Therefore, the protection scope of the patent of this application shall be subject to the appended claims.

Claims

1. A user preference analysis method, characterized in that: The analysis method comprises: Input the heat value sequence of each demand of the target user into the long short-term memory network to obtain the long-term and short-term features of each demand. The long-term features of each demand constitute the user profile of the target user. Cluster the user portraits of all users to obtain the user group to which the target user belongs; Obtain a correlation matrix of the user group, and input the short-term characteristics and correlation matrix of each demand into the preference analysis model to obtain the heat decay coefficient of each demand, wherein the correlation matrix includes the correlation of heat value changes between any demands; Input the real-time popularity value and popularity decay coefficient of each demand of the target user into the preset decay model to predict the popularity duration of each demand, and take the popularity duration of each demand as the analysis result; The training method of the preference analysis model includes: obtaining the short-term characteristics of each demand and the correlation matrix of the user group to which the user belongs according to the first historical sequence of the heat value of each demand of any user as a set of training samples; fitting a preset attenuation model in the second historical sequence of the heat value of each demand after the first historical sequence of the heat value of the user to obtain the attenuation coefficient label of the training sample; inputting the training sample into the preference analysis model to obtain the output result, calculating the mean square error loss function of the output result and the attenuation coefficient label to iteratively train the preference analysis model; need and demand The calculation method of the correlation of the heat value changes between the requirements includes: for any user in the user group, calculating the heat value difference between the current collection cycle and the previous adjacent collection cycle, obtaining the heat value change sequence of each requirement, and calculating the heat value difference between the user in the requirement and demand The Pearson correlation coefficient of the heat value change sequence between them; calculate the representativeness of the user, which is negatively correlated with the Euclidean distance between the user portrait and the user group cluster center; perform weighted summation of the Pearson correlation coefficients of all users according to the representativeness to obtain the demand and demand The correlation between the changes in heat values.

2. A user preference analysis method according to claim 1, characterized in that: The method for obtaining the heat value includes: Obtaining multiple behavioral data of target users for any needs within a collection cycle, the behavioral data including the number of clicks, browsing time and purchase times; The behavior score of each behavior data is determined according to a preset score range, and each behavior score is weighted and summed according to a preset weight to obtain the heat value of the target user for the demand during the collection period.

3. A user preference analysis method according to claim 1, characterized in that: Cluster the user portraits of all users to obtain the user group to which the target user belongs, including: The Kmeans algorithm is used to cluster user portraits, and the elbow method is used to determine the number of clusters, and multiple clusters and the cluster centers of each cluster are obtained; The Euclidean distance between the user profile of the target user and each cluster center is calculated, and the cluster to which the cluster center corresponding to the minimum Euclidean distance belongs is taken as the user group to which the target user belongs.

4. A user preference analysis method according to claim 1, characterized in that: user Representativeness for: , For users in the user group User portraits, is the cluster center of the user group, for and The Euclidean distance of User portraits and The sum of the Euclidean distances.

5. A user preference analysis method according to claim 1, characterized in that: The preference analysis model includes a convolution sub-model and a regression sub-model; the convolution sub-model is used to extract features from the correlation matrix to obtain correlation features; the correlation features and the short-term features of each demand are spliced ​​and input into the regression sub-model to output the heat attenuation coefficient of each demand.

6. A user preference analysis method according to claim 1, characterized in that: The predicted heat duration of each demand includes: Input the real-time heat value and heat attenuation coefficient of any demand of the target user into a preset attenuation model to obtain the predicted heat value of the demand in multiple future collection cycles; The first future collection cycle in which the predicted heat value is less than the heat threshold is taken as the cutoff cycle, and the number of collection cycles between the current collection cycle and the cutoff cycle is taken as the heat duration of the demand.

7. A user preference analysis method according to claim 1, characterized in that: The preset attenuation model is: , is the real-time heat value, is the heat attenuation coefficient, For the A collection cycle, For the The predicted heat value of a collection cycle.

8. A user preference analysis method according to claim 1, characterized in that: The calculation method of the mean square error loss function includes: Calculate the average variance of the heat value of each demand in the user group to which the user belongs; mean square error loss function satisfy: , is the quantity of all requirements, Required in the attenuation coefficient label The thermal attenuation coefficient, For the output results The thermal attenuation coefficient, For demand The average variance of the heat value, is the sum of the average variance of the heat values ​​of all requirements, It is the sum of the absolute values ​​of the heat attenuation coefficients in the output results.

9. A user preference analysis system, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a user preference analysis method according to any one of claims 1 to 8 is implemented.

Images