Intelligent recommendation method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention discloses an intelligent recommendation method and device, electronic equipment and a storage medium. Obtaining the access position of the user to be recommended and the product access rate of the candidate product; determining a target area according to the access position, and obtaining the product access rate of each candidate product by each user in the same area in the target area; according to the product access rate of the to-be-recommended user to each candidate product and the product access rate of each user in the same area to each candidate product, carrying out clustering to obtain a preference matrix; determining the user similarity between the to-be-recommended user and each candidate similar user in the preference matrix according to the preference matrix; and according to the user similarity and the preference matrix, determining a recommended product. According to the embodiment of the invention, the recommendation accuracy is improved.

Description

Technical Field

[0001] The embodiments of the present application relate to artificial intelligence technology, and in particular to an intelligent recommendation method, device, electronic device and storage medium. Background Art

[0002] With the rapid development of the Internet, corresponding products are constantly updated, product information on the Internet is expanding, and the phenomenon of information overload is becoming more and more prominent. It is difficult for users to find the product information they need from the massive information resources.

[0003] In the prior art, recommendation algorithms are usually used to analyze different users, items, and interaction information between users and items, determine user preferences, and recommend items that best meet user needs.

[0004] However, the accuracy of item recommendations is affected by the quality of the interactive information data between users and items, resulting in low recommendation accuracy. Summary of the invention

[0005] The present application provides an intelligent recommendation method, device, electronic device and storage medium to improve the accuracy of recommendations.

[0006] In a first aspect, an embodiment of the present application provides an intelligent recommendation method, the intelligent recommendation method comprising:

[0007] Obtain the access location of the user to be recommended and the product access rate of the candidate product;

[0008] Determine the target area based on the access location, and obtain the product access rate of each candidate product by users in the same area of ​​the target area;

[0009] According to the product visit rate of each candidate product by the users to be recommended and the product visit rate of each candidate product by users in the same region, a preference matrix is ​​obtained;

[0010] Determine the user similarity between the user to be recommended and each candidate similar user in the preference matrix according to the preference matrix;

[0011] Determine recommended products based on user similarity and preference matrix.

[0012] In a second aspect, an embodiment of the present application further provides an intelligent recommendation device, the intelligent recommendation device comprising:

[0013] A user information acquisition module is used to obtain the access location of the user to be recommended and the product access rate of the candidate product;

[0014] A target area determination module is used to determine the target area according to the access location and obtain the product access rate of each candidate product by users in the same area of ​​the target area;

[0015] A preference matrix determination module is used to cluster the product access rates of each candidate product by the user to be recommended and the product access rates of each candidate product by users in the same region to obtain a preference matrix;

[0016] A user similarity determination module is used to determine the user similarity between the user to be recommended and each candidate similar user in the preference matrix according to the preference matrix;

[0017] The recommended product determination module is used to determine the recommended products based on user similarity and preference matrix.

[0018] In a third aspect, an embodiment of the present application further provides an electronic device, the electronic device comprising:

[0019] one or more processors;

[0020] A storage device for storing one or more programs;

[0021] When one or more programs are executed by one or more processors, the one or more processors implement any one of the intelligent recommendation methods provided in the embodiments of the present application.

[0022] In a fourth aspect, an embodiment of the present application further provides a storage medium comprising computer executable instructions, which, when executed by a computer processor, are used to execute any one of the intelligent recommendation methods provided in the embodiments of the present application.

[0023] The present application obtains the access location of the user to be recommended and the product access rate of the candidate products; determines the target area according to the access location, and obtains the product access rate of each user in the same area of ​​the target area to each candidate product, and improves the accuracy of subsequent user similarity through regional restrictions; clusters the product access rate of each candidate product to the user to be recommended and the product access rate of each candidate product to the user in the same area to obtain a preference matrix, and overcomes the data sparsity problem of the preference matrix in the prior art through clustering, and improves the data quality of the interactive information between users and items; determines the user similarity between the user to be recommended and each candidate similar user in the preference matrix according to the preference matrix; determines the recommended product according to the user similarity and the preference matrix, and improves the accuracy of the recommended product based on the high quality of the data in the preference matrix after clustering. Therefore, through the technical solution of the present application, the problem that the accuracy of recommendation is affected by the quality of the interactive information data between users and items, and the low accuracy of recommendation is solved, and the effect of improving the accuracy of recommendation is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a flow chart of an intelligent recommendation method in Embodiment 1 of the present application;

[0025] Figure 2This is a flow chart of an intelligent recommendation method in Embodiment 2 of the present application;

[0026] Figure 3 This is a flow chart of an intelligent recommendation method in Embodiment 3 of the present application;

[0027] Figure 4 This is a flow chart of an intelligent recommendation method in Embodiment 4 of the present application;

[0028] Figure 5 It is a structural diagram of an intelligent recommendation device in Embodiment 5 of the present application;

[0029] Figure 6 It is a structural diagram of an electronic device in Example 6 of the present application. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings 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 should fall within the scope of protection of the present application.

[0031] It should be noted that the terms "first" and "second" etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0032] Embodiment 1

[0033] Figure 1 This is a flowchart of an intelligent recommendation method provided in Example 1 of the present application. This embodiment can be applied to situations where product recommendations are made to users. The method can be executed by an intelligent recommendation device, which can be implemented in software and / or hardware and specifically configured in an electronic device that requires an application recommendation function, such as a search engine.

[0034] See also Figure 1 The intelligent recommendation method shown specifically includes the following steps:

[0035] S110: Obtain the access location of the user to be recommended and the product access rate of the candidate product.

[0036] The access location may be the geographical location of the user when accessing the application for which intelligent recommendation is required, and is used to determine the recommended products for the user. Exemplarily, the user's access location may be obtained through positioning software. The access location may be a certain access location, or may be a frequently used location determined through multiple access locations of the user. For example, the access location may be a location determined by clustering multiple access locations of the user through a clustering algorithm.

[0037] The product access rate may be the access frequency of all candidate products in the visited application, and is used to determine recommended products for the user. Exemplarily, the user's access probability may be obtained through the user's historical access log, and the user's access data may be stored in a special database, and the product access rate of the candidate products may be obtained from the database. The candidate products may be products in the visited application. Exemplarily, the candidate products may be articles and electronic services, etc., which are not specifically limited in this application.

[0038] S120: Determine a target area according to the access location, and obtain product access rates of users in the target area to each candidate product.

[0039] The target area may be an area determined according to the access location and the preset area range, and is used to determine users in the same area. The preset area range may be a preset range of the target area determined according to the access location, and is used to determine the target area. For example, the preset area range may be a circular range with the access location as the center and a preset value as the radius.

[0040] The same-region user may be a user whose access location is within the target region, and is used to determine the preference similarity between the to-be-recommended user and each same-region user. Exemplarily, the product access rate of each same-region user to each candidate product may be obtained by reading from a corresponding database based on the identifier of each same-region user.

[0041] S130 , clustering is performed according to the product access rates of the to-be-recommended users to each candidate product and the product access rates of the same-region users to each candidate product to obtain a preference matrix.

[0042] According to the product access rate of the user to be recommended to each candidate product, and the product access rate of each user in the same region to each candidate product, a preference matrix can be obtained through a clustering algorithm. The preference matrix can be a matrix including the user to be recommended and the interest of the user in the same region who is similar to the product of interest to the user to be recommended, and is used to determine the user similarity between the user to be recommended and each candidate similar user in the preference matrix. Exemplarily, the clustering algorithm can be a K-means (professional term, a clustering algorithm) algorithm.

[0043] Clustering refers to the task of dividing data into multiple groups, each of which can be called a cluster. The goal of clustering is to divide data so that the elements in each group are very similar, but the data in different groups are very different, that is, classification. Clustering can be used to classify similar data into one category, making the data category clearer. K-means (a professional term, a clustering algorithm) is the most typical clustering algorithm, and it is also one of the simplest and most commonly used algorithms. The main function of this algorithm is to automatically classify similar samples into one category. By setting a reasonable K value (a professional term, a parameter in the K-means clustering algorithm), different clustering effects can be determined.

[0044] Product access rate can be used to represent user preferences, but a single user may frequently use a single business function, which cannot fully represent the user's needs for the business, and there is a data sparsity problem. Therefore, clustering algorithms can be used to divide users with similar preferences into the same cluster and build a dense preference matrix to solve the data sparsity problem.

[0045] S140: Determine the user similarity between the user to be recommended and each candidate similar user in the preference matrix according to the preference matrix.

[0046] The candidate similar users may be users other than the user to be recommended in the preference matrix. The user similarity may be the similarity between the user to be recommended and the products of interest of each candidate similar user in the preference matrix, determined based on the preference matrix and a preset similarity algorithm. Exemplarily, the user similarity may be determined based on the similarity between the user to be recommended and the candidate similar users in the preference matrix. The preset similarity algorithm may be a pre-set user similarity determination algorithm. For example, the user similarity algorithm may be a Pearson correlation coefficient.

[0047] S150: Determine recommended products based on user similarity and preference matrix.

[0048] The recommended product may be a product recommended to the user to be recommended based on the user similarity and preference matrix. For example, based on the user similarity, a candidate similar user with a high similarity to the user to be recommended may be determined as a target similar user, and based on the preference matrix, a product with a high degree of interest to the target similar user may be determined as a recommended product.

[0049] With the rapid development of the Internet, the amount of information on the Internet is expanding, and the phenomenon of information overload is becoming more and more prominent. It is difficult for users to find the information they need from the massive amount of information resources. For the financial industry, with the increase in the categories of financial products, as well as various physical products and service products launched by banks for users, users find it increasingly difficult to obtain the information they want from the numerous products. They may even need to search on the search page or ask customer service to find the products they need.

[0050] In order to solve the above problems, recommendation systems have gradually emerged in people's vision. In recommendation systems, it is not necessary to know in advance what information users need, but to directly find the information that users are interested in from the massive amount of information and recommend it to users. Recommendation systems are proactive and can make up for the shortcomings of search engines. Recommendation systems analyze different users, different products, and the interactive information between them to determine user preferences and recommend products that best meet user needs.

[0051] In practical applications, search engines and recommendation systems are usually combined for user recommendations. Search engines can solve the most direct needs of users, while recommendation systems can discover the implicit needs of users. The two systems work together to meet the personalized needs of users. Recommendation systems enable effective information exchange between information resources and users, greatly improving the efficiency of information utilization.

[0052] In the recommendation system, the recommendation algorithm is the most important part. The quality of the recommendation algorithm determines the quality of the recommendation to a certain extent. Applying personalized recommendation algorithms to the recommendation system can effectively screen and filter information by mining the relationship between users and products, retrieve corresponding information resources according to users' interests and needs, and recommend related products, alleviating the problem of information overload.

[0053] Among the many recommendation algorithms, collaborative filtering algorithm is the most popular and widely used recommendation technology, which has significant advantages over traditional recommendation algorithms. Collaborative filtering recommendation algorithm is a recommendation algorithm based on user behavior analysis, which requires users to continuously interact with the system and generate a large amount of behavior data. It collects a large amount of data by analyzing the user's feedback information on each item, and recommends products that meet user needs to users.

[0054] However, collaborative filtering recommendation algorithms still face problems with data sparsity and recommendation accuracy. Data sparsity has always been one of the difficulties that hinder the application of collaborative filtering recommendation algorithms. Data sparsity is usually caused by insufficient or even missing useful information. When the input data structure is tilted or unevenly distributed in other ways, the interactive data of users' access rate to products and the overall preference of users in the region for products are relatively limited, which makes the preference matrix often very sparse. Since there is not enough data for recommendation within the recommendation system, a large part of users cannot obtain information resources that meet their needs, and many products do not even have the opportunity to appear in the recommendation results.

[0055] The technical solution of this embodiment is to obtain the access location of the user to be recommended and the product access rate of the candidate products; determine the target area according to the access location, and obtain the product access rate of each user in the same area of ​​the target area to each candidate product, and improve the accuracy of subsequent user similarity through regional restrictions; cluster the product access rate of the user to be recommended to each candidate product and the product access rate of each user in the same area to each candidate product to obtain a preference matrix, and overcome the data sparsity problem of the preference matrix in the prior art through clustering, and improve the data quality of the interactive information between users and items; determine the user similarity between the user to be recommended and each candidate similar user in the preference matrix according to the preference matrix; determine the recommended product according to the user similarity and the preference matrix, and improve the accuracy of the recommended product based on the high quality of the data in the preference matrix after clustering. Therefore, through the technical solution of this application, the problem that the accuracy of recommendation is affected by the quality of the interactive information data between users and items, and the low accuracy of recommendation is solved, and the effect of improving the accuracy of recommendation is achieved.

[0056] Embodiment 2

[0057] Figure 2 This is a flow chart of an intelligent recommendation method provided in Example 2 of the present application. The technical solution of this embodiment is further refined on the basis of the above technical solution.

[0058] Furthermore, "clustering according to the product visit rate of each candidate product by the users to be recommended, and the product visit rate of each candidate product by users in the same region to obtain a preference matrix" is refined as: "determining the initial preference matrix according to the product visit rate of each candidate product by the users to be recommended, the product visit rate of each candidate product by users in the same region, and the product affiliation of each candidate product; updating the initial preference matrix through a clustering algorithm to obtain a preference matrix" to obtain the preference matrix.

[0059] See also Figure 2 An intelligent recommendation method shown includes:

[0060] S210: Obtain the access location of the user to be recommended and the product access rate of the candidate product.

[0061] S220: Determine a target area according to the access location, and obtain product access rates of users in the target area to each candidate product.

[0062] S230: Determine an initial preference matrix according to the product access rate of each candidate product by the user to be recommended, the product access rate of each candidate product by users in the same region, and the product affiliation of each candidate product.

[0063] According to the product access rate of each candidate product by the user to be recommended and the product access rate of each candidate product by users in the same region, the access rate can be used to represent the global preference of users in the region. The product access rate of each candidate product by different users can be saved in dictionary form. In reality, not all candidate products will be accessed by different users. If the product access rate of each candidate product by different users is stored in vector form, it is generally a sparse array, or there will be a large number of null values. Therefore, the product access rate of each candidate product by different users is saved in dictionary form. The data type is a dictionary, the key is different users, and the value is also a dictionary, which stores the user's access rate to different services. Exemplarily, the dictionary storage form of the product access rate of each candidate product by different users is as follows:

[0064] ﹛

[0065] "User 1": {"Product 1"; 5,"Product 2"; 5,"Product 3"; 3,"Product 4"; 4,"Product 5"; 0},

[0066] "User 2": {"Product 1"; 4,"Product 2"; 5,"Product 3"; 3,"Product 4"; 8,"Product 5"; 2},

[0067] "User 3": {"Product 1"; 2,"Product 2"; 7,"Product 3"; 6,"Product 4"; 5,"Product 5"; 3},

[0068] "User 4": {"Product 1"; 2,"Product 2"; 3,"Product 3"; 4,"Product 4"; 4,"Product 5"; 6},

[0069] "User5":{"Product1";2,"Product2";5,"Product3";5,"Product4";7,"Product5";4}

[0070] ﹜

[0071] According to the product access rate of each candidate product by the user to be recommended and the product access rate of each candidate product by the users in the same region, a access rate matrix can be constructed. For example, the access rate matrix is ​​shown as follows:

[0072]

[0073] Among them, r nm Represents the access rate of the nth user to the mth candidate product.

[0074] The product affiliation of each candidate product may be a category affiliation of the candidate product. A category affiliation matrix is ​​constructed according to the product category affiliation of the candidate product. Exemplarily, the category affiliation matrix is ​​shown as follows:

[0075]

[0076] Among them, f mk Indicates whether the mth candidate product belongs to the kth product category.

[0077] Products are large categories, and categories refer to small categories. For example, the service product transfer and remittance in mobile banking is a large category, and the small categories are the registered account transfer, domestic remittance, cross-border remittance and other small categories included in the transfer and remittance. For example, the value of each element in the category affiliation matrix can be determined by the following formula:

[0078]

[0079] According to the category membership matrix and the access rate matrix, an initial preference matrix can be constructed. For example, according to the membership relationship and access rate of each candidate product, the formula for determining the preference is as follows:

[0080]

[0081] Among them, r u =(r u1 ,r u2 ,r u3 ,...,r um ) is the access rate vector of user u to each candidate product, fi=(f 1i ,f 2i ,f 3i ,...,f mi ) is the user vector of users in the same region corresponding to candidate product i.

[0082] Exemplarily, the initial preference matrix is ​​shown as follows:

[0083]

[0084] The access rate matrix is ​​usually a sparse matrix because the number of users and products is extremely large, while the number of products associated with a single user is extremely small. The value of k in the category membership matrix is ​​smaller than the number of products m in the access rate matrix. Therefore, the initial preference matrix constructed by the category membership matrix and the access rate matrix is ​​reduced in dimension relative to the access rate matrix, which is conducive to reducing the time and space complexity of determining the recommended products later.

[0085] S240. Update the initial preference matrix by using a clustering algorithm to obtain a preference matrix.

[0086] Before updating the initial preference matrix through the clustering algorithm, the initial preference matrix is ​​normalized. The normalization process regulates the numerical range in the initial preference matrix and reduces the amount of subsequent data calculation. Exemplarily, the normalization formula is shown as follows:

[0087]

[0088] Among them, x i,j is the value corresponding to the i-th row and j-th column of the initial preference matrix, which represents the interest of user i in service j in the initial preference matrix. min is the minimum value of all users’ interest in the product, x max is the maximum value of all users’ interest in the product.

[0089] Exemplarily, the clustering algorithm may be a fuzzy C-means clustering algorithm using particle swarm optimization. The steps of updating the initial preference matrix by the clustering algorithm to obtain the preference matrix may be as follows:

[0090] Step 1: Set the number of categories C and the fuzzy coefficient m. C and m can be determined by professional technicians based on experience or experiments, and this application does not make specific restrictions on this. For example, m can be 2;

[0091] Step 2: Obtain the normalized initial preference matrix;

[0092] Step 3: Introduce the Lagrangian factor to construct the objective function:

[0093]

[0094] Here d(x i ,c i ) represents the relationship between the ith data point and the cth data point i The Euclidean distance between cluster centers, u ij is the element in the normalized initial preference matrix, c i is the cluster center corresponding to each cluster;

[0095] Step 4: Set the optimization conditions for finding the extreme value of the objective function as follows:

[0096]

[0097] Step 5: Based on the optimization conditions of step 4, we get the formula:

[0098]

[0099] Step 6: Calculate the cluster center according to step 5;

[0100] Step 7: Update the initial preference matrix according to step 4;

[0101] Step 8: Termination procedure: Compare the initial preference matrices of the iteration according to the matrix paradigm. If ||U (t) -U (t-1) ||<ε, the iteration stops, otherwise return to step 6.

[0102] S250: Determine the user similarity between the user to be recommended and each candidate similar user in the preference matrix according to the preference matrix.

[0103] S260: Determine recommended products based on user similarity and preference matrix.

[0104] The technical solution of this embodiment determines the initial preference matrix based on the product access rate of each candidate product by the users to be recommended, the product access rate of each candidate product by users in the same area, and the product affiliation of each candidate product, thereby overcoming the problem of data sparsity of the access rate matrix and reducing the time and space complexity of subsequent determination of recommended products; the initial preference matrix is ​​updated by a clustering algorithm to obtain the preference matrix, and the membership of each data point in the preference matrix to all class centers is obtained by optimizing the objective function, thereby determining the category of the data point to achieve the purpose of automatically classifying the data points in the preference matrix, thereby improving the similarity of users in the preference matrix and improving the accuracy of subsequent recommendations.

[0105] Embodiment 3

[0106] Figure 3 This is a flow chart of an intelligent recommendation method provided in Example 3 of the present application. The technical solution of this embodiment is further refined on the basis of the above technical solution.

[0107] Furthermore, "determine the user similarity between the user to be recommended and each candidate similar user in the preference matrix based on the preference matrix" is refined as: "determine the Pearson correlation coefficient between the user to be recommended and each candidate similar user based on the interest of the user to be recommended and each candidate similar user in each candidate interest product in the preference matrix; use the Pearson correlation coefficient between the user to be recommended and each candidate similar user as the user similarity between the user to be recommended and each candidate similar user" to determine the user similarity between the user to be recommended and each candidate similar user.

[0108] See also Figure 3 An intelligent recommendation method shown includes:

[0109] S310: Obtain the access location of the user to be recommended and the product access rate of the candidate product.

[0110] S320: Determine a target area according to the access location, and obtain product access rates of users in the target area to each candidate product.

[0111] S330 , clustering is performed according to the product access rates of the to-be-recommended users to each candidate product and the product access rates of the same-region users to each candidate product to obtain a preference matrix.

[0112] S340 , determining the Pearson correlation coefficient between the user to be recommended and each candidate similar user according to the interest of the user to be recommended and each candidate similar user in the preference matrix for each candidate interest product.

[0113] The Pearson correlation coefficient can be a statistic used to measure the degree of linear correlation between two variables. According to the interest of the user to be recommended and each candidate similar user in each candidate interest product in the preference matrix, the formula for determining the Pearson correlation coefficient between the user to be recommended and each candidate similar user is as shown in the following formula:

[0114]

[0115] Among them, I u,v represents the set of features that users u and v prefer, r u,i is the preference of user u for feature i, r v,i is the preference of user v for feature i, represents the average preference of user u for all candidate recommended products, It represents the average preference of user v for all candidate recommended products.

[0116] S350: Use the Pearson correlation coefficient between the user to be recommended and each candidate similar user as the user similarity between the user to be recommended and each candidate similar user.

[0117] The Pearson correlation coefficient between the user to be recommended and each candidate similar user is used as the user similarity between the user to be recommended and each candidate similar user, thereby improving the accuracy of the user similarity.

[0118] S360. Determine recommended products based on user similarity and preference matrix.

[0119] The technical solution of this embodiment determines the Pearson correlation coefficient between the user to be recommended and each candidate similar user according to the interest of the user to be recommended and each candidate similar user in each candidate interest product in the preference matrix; the Pearson correlation coefficient between the user to be recommended and each candidate similar user is used as the user similarity between the user to be recommended and each candidate similar user, and the Pearson correlation coefficient is determined based on the feature set of common preferences of the user to be recommended and each candidate similar user, thereby improving the accuracy of user similarity.

[0120] Embodiment 4

[0121] Figure 4 This is a flow chart of an intelligent recommendation method provided in Example 4 of the present application. The technical solution of this embodiment is further refined on the basis of the above technical solution.

[0122] Furthermore, "determine recommended products based on user similarity and preference matrix" is refined as: "determine the predicted interest of the recommended user in each candidate interest product in the preference matrix based on user similarity and preference matrix; determine recommended products based on the predicted interest of each candidate interest product" to determine the recommended products.

[0123] See also Figure 4 An intelligent recommendation method shown includes:

[0124] S410: Obtain the access location of the user to be recommended and the product access rate of the candidate product.

[0125] S420: Determine a target area according to the access location, and obtain product access rates of users in the target area to each candidate product.

[0126] S430 , clustering is performed according to the product access rates of the to-be-recommended users to each candidate product and the product access rates of the same-region users to each candidate product to obtain a preference matrix.

[0127] S440: Determine the user similarity between the user to be recommended and each candidate similar user in the preference matrix according to the preference matrix.

[0128] S450: Determine the predicted interest of the user to be recommended for each candidate product of interest in the preference matrix according to the user similarity and the preference matrix.

[0129] The candidate products of interest may be products in the preference matrix, which are used to determine the recommended products. The predicted interest may be a predicted value of the interest of the user to be recommended in the candidate products of interest, which is used to determine the recommended products. The predicted interest of the user to be recommended in each candidate product of interest in the preference matrix is ​​determined based on the interest of the candidate similar users in the preference matrix in each candidate product of interest. Exemplarily, the average or median of the interest of the candidate similar users with a high user similarity to the user to be recommended in the candidate products of interest may be used as the predicted interest of the user to be recommended in each candidate product of interest in the preference matrix.

[0130] In an optional embodiment, based on user similarity and a preference matrix, the predicted interest of the user to be recommended in each candidate product of interest in the preference matrix is ​​determined, including: determining the target similar users of the user to be recommended based on user similarity and a preset number of similar users; and determining the predicted interest of the user to be recommended in each candidate product of interest based on the target similar users and the preference matrix.

[0131] The preset number of similar users may be the preset number of target similar users. The preset number of similar users may be determined by professional technicians based on experience or experiments, and this application does not specifically limit this. The target similar users may be candidate similar users within the preset number of similar users after sorting the user similarities from large to small, and are used to determine the predicted interest of the user to be recommended in each candidate product of interest. By presetting the number of similar users, determining the target similar users of the user to be recommended, and reducing the interference of users with lower similarity, the accuracy of subsequent recommended products can be improved.

[0132] According to the interest of the target similar users in the preference matrix for the candidate interest products that overlap with the user to be recommended, the predicted interest of the user to be recommended for each candidate interest product is determined. For example, the average, expected or median of the interest of the target similar users in the preference matrix for the candidate interest products that overlap with the user to be recommended can be determined as the predicted interest of the user to be recommended for each candidate interest product.

[0133] By determining the target similar users of the recommended users based on user similarity and the preset number of similar users and eliminating the interference of users with lower similarity, the accuracy of subsequent predicted interest can be improved; based on the target similar users and the preference matrix, the predicted interest of the recommended users in each candidate product of interest is determined to improve the accuracy of predicted interest, thereby improving the accuracy of recommended products.

[0134] In an optional embodiment, the predicted interest of the user to be recommended in each candidate product of interest is determined based on the target similar users and the preference matrix, including: based on the target similar users and the preference matrix, the preference of each target similar user for each common product in the set of products visited by the user to be recommended, the average preference of the user to be recommended for each candidate product of interest, and the average preference of each target similar user for each candidate product of interest, determining the predicted visit rate of the user to be recommended for each candidate product of interest.

[0135] Exemplarily, the predicted access rate of each candidate product by the user to be recommended may be determined by the following formula:

[0136]

[0137] Among them, S u represents the target similar user set; I u,v represents the set of candidate products of interest that users u and v visit together; Sim(u,v) is the user similarity between users u and v, represents the average preference of user u for all candidate recommended products, It represents the average preference of user v for all candidate recommended products.

[0138] By determining the predicted visit rate of the user to be recommended to each candidate product of interest based on the target similar users and the preference matrix, the preference of each target similar user for each common product in the product set visited by the user to be recommended, the average preference of the user to be recommended to each candidate product of interest, and the average preference of each target similar user to each candidate product of interest, the predicted visit rate of the user to be recommended to each candidate product of interest is determined by considering the preference of each target similar user for each common product in the product set visited by the user to be recommended, and the average preference of the target similar users and the user to be recommended for all candidate recommended products, the accuracy of the predicted visit rate of the recommended user to each candidate product of interest is improved, and the accuracy of the recommended products is improved.

[0139] S460: Determine recommended products based on the predicted interest levels of the candidate interest products.

[0140] The predicted interest of each candidate product of interest is sorted from large to small, and according to the preset number of recommendations, the first recommended number of candidate products of interest are recommended as recommended products to the user to be recommended. The number of recommendations can be determined according to the settings of the recommendation system and user habits, and this application does not make specific restrictions on this.

[0141] The technical solution of this embodiment determines the predicted interest of the recommended user for each candidate interest product in the preference matrix based on the user similarity and the preference matrix, and determines the predicted interest based on the user similarity and the preference matrix to improve the accuracy of the predicted interest; determines the recommended product based on the predicted interest of each candidate interest product to improve the accuracy of the recommended product.

[0142] Embodiment 5

[0143] Figure 5 The figure shows a schematic diagram of the structure of an intelligent recommendation device provided in the fifth embodiment of the present application. This embodiment is applicable to the case of recommending products to users. The specific structure of the intelligent recommendation device is as follows:

[0144] The user information acquisition module 510 is used to obtain the access location of the user to be recommended and the product access rate of the candidate product;

[0145] The target area determination module 520 is used to determine the target area according to the access location, and obtain the product access rate of each user in the same area of ​​the target area to each candidate product;

[0146] The preference matrix determination module 530 is used to cluster the product access rates of the recommended users to each candidate product and the product access rates of the same-region users to each candidate product to obtain a preference matrix;

[0147] A user similarity determination module 540 is used to determine the user similarity between the user to be recommended and each candidate similar user in the preference matrix according to the preference matrix;

[0148] The recommended product determination module 550 is used to determine recommended products according to user similarity and preference matrix.

[0149] The technical solution of this embodiment is to obtain the access location of the user to be recommended and the product access rate of the candidate products; determine the target area according to the access location, and obtain the product access rate of each user in the same area of ​​the target area to each candidate product, and improve the accuracy of subsequent user similarity through regional restrictions; cluster the product access rate of the user to be recommended to each candidate product and the product access rate of each user in the same area to each candidate product to obtain a preference matrix, and overcome the data sparsity problem of the preference matrix in the prior art through clustering, and improve the data quality of the interactive information between users and items; determine the user similarity between the user to be recommended and each candidate similar user in the preference matrix according to the preference matrix; determine the recommended product according to the user similarity and the preference matrix, and improve the accuracy of the recommended product based on the high quality of the data in the preference matrix after clustering. Therefore, through the technical solution of this application, the problem that the accuracy of recommendation is affected by the quality of the interactive information data between users and items, and the low accuracy of recommendation is solved, and the effect of improving the accuracy of recommendation is achieved.

[0150] Optionally, the preference matrix determination module 530 includes:

[0151] An initial preference matrix determination unit, used to determine an initial preference matrix according to the product access rate of each candidate product by the user to be recommended, the product access rate of each candidate product by users in the same region, and the product affiliation of each candidate product;

[0152] The initial preference matrix updating unit is used to update the initial preference matrix through a clustering algorithm to obtain a preference matrix.

[0153] Optionally, the user similarity determination module 540 includes:

[0154] A Pearson correlation coefficient determination unit, used to determine the Pearson correlation coefficient between the user to be recommended and each candidate similar user according to the interest of the user to be recommended and each candidate similar user in each candidate interest product in the preference matrix;

[0155] The user similarity determination unit is used to use the Pearson correlation coefficient between the user to be recommended and each candidate similar user as the user similarity between the user to be recommended and each candidate similar user.

[0156] Optionally, the recommended product determination module 550 includes:

[0157] A predicted interest determination unit, used to determine the predicted interest of the user to be recommended for each candidate product of interest in the preference matrix according to the user similarity and the preference matrix;

[0158] The recommended product determination unit is used to determine the recommended product according to the predicted interest level of each candidate interest product.

[0159] Optionally, the predicted interest determination unit includes:

[0160] A target similar user determination subunit is used to determine target similar users of the user to be recommended based on user similarity and a preset number of similar users;

[0161] The predicted interest level determination subunit is used to determine the predicted interest level of the user to be recommended to each candidate product of interest based on the target similar users and the preference matrix.

[0162] Optionally, the predicted interest level determination subunit is specifically used for:

[0163] According to the target similar users and preference matrix, the preference of each target similar user for each common product in the common visited product set with the user to be recommended, the average preference of the user to be recommended for each candidate interest product, and the average preference of each target similar user for each candidate interest product, the predicted visit rate of the user to be recommended for each candidate interest product is determined.

[0164] The intelligent recommendation device provided in the embodiments of the present application can execute the intelligent recommendation method provided in any embodiment of the present application, and has the corresponding functional modules and beneficial effects for executing the intelligent recommendation method.

[0165] According to an embodiment of the present invention, the present invention also provides an electronic device, a readable storage medium and a computer program product.

[0166] Embodiment 6

[0167] Figure 6 A schematic diagram of the structure of an electronic device provided in Example 6 of the present application is shown in FIG. Figure 6 As shown, the electronic device includes a processor 610, a memory 620, an input device 630, and an output device 640; the number of the processor 610 in the electronic device can be one or more. Figure 6 A processor 610 is taken as an example; the processor 610, the memory 620, the input device 630 and the output device 640 in the electronic device can be connected via a bus or other means. Figure 6 The example of connecting through bus is taken in the following.

[0168] The memory 620, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the intelligent recommendation method in the embodiment of the present application (for example, user information acquisition module 510, target area determination module 520, preference matrix determination module 530, user similarity determination module 540 and recommended product determination module 550). The processor 610 executes various functional applications and data processing of the electronic device by running the software programs, instructions and modules stored in the memory 620, that is, realizing the above-mentioned intelligent recommendation method.

[0169] The memory 620 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 620 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 620 may further include a memory remotely arranged relative to the processor 610, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0170] The input device 630 may be used to receive input character information and generate key signal input related to user settings and function control of the electronic device. The output device 640 may include a display device such as a display screen.

[0171] Embodiment 7

[0172] Embodiment 7 of the present application also provides a storage medium containing computer executable instructions, which are used to execute an intelligent recommendation method when executed by a computer processor. The method includes: obtaining the access location of the user to be recommended and the product access rate of the candidate products; determining the target area according to the access location, and obtaining the product access rate of each user in the same area within the target area to each candidate product; clustering the product access rate of each candidate product by the user to be recommended and the product access rate of each user in the same area to each candidate product to obtain a preference matrix; determining the user similarity between the user to be recommended and each candidate similar user in the preference matrix according to the preference matrix; determining the recommended product according to the user similarity and the preference matrix.

[0173] Of course, the storage medium containing computer executable instructions provided in the embodiment of the present application, whose computer executable instructions are not limited to the above method operations, can also execute related operations in the intelligent recommendation method provided in any embodiment of the present application.

[0174] Through the above description of the implementation method, the technicians in the relevant field can clearly understand that the present application can be implemented with the help of software and necessary general hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application can be essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for an electronic device (which can be a personal computer, server, or network device, etc.) to execute the methods of each embodiment of the present application.

[0175] It is worth noting that in the embodiment of the above-mentioned intelligent recommendation device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application.

[0176] Note that the above are only preferred embodiments of the present application and the technical principles used. Those skilled in the art will understand that the present application is not limited to the specific embodiments herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present application. Therefore, although the present application is described in more detail through the above embodiments, the present application is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. An intelligent recommendation method, characterized in that: include: Obtain the access location of the user to be recommended and the product access rate of the candidate product; Determine a target area according to the access location, and obtain product access rates of users in the target area to each of the candidate products; Clustering is performed according to the product access rates of the to-be-recommended users to each of the candidate products and the product access rates of the users in the same region to each of the candidate products to obtain a preference matrix; Determining the user similarity between the user to be recommended and each candidate similar user in the preference matrix according to the preference matrix; Determine recommended products based on the user similarity and the preference matrix.

2. The method according to claim 1, characterized in that The preference matrix is ​​obtained by clustering the product access rates of the to-be-recommended users to the candidate products and the product access rates of the same-region users to the candidate products, including: Determine an initial preference matrix according to the product access rates of the to-be-recommended users to the candidate products, the product access rates of the same-region users to the candidate products, and the product affiliations of the candidate products; The initial preference matrix is ​​updated by a clustering algorithm to obtain a preference matrix.

3. The method according to claim 1, characterized in that The determining, according to the preference matrix, the user similarity between the to-be-recommended user and each candidate similar user in the preference matrix comprises: Determine the Pearson correlation coefficient between the user to be recommended and each of the candidate similar users according to the interest of the user to be recommended and each of the candidate similar users in the preference matrix for each candidate product of interest; The Pearson correlation coefficient between the user to be recommended and each of the candidate similar users is used as the user similarity between the user to be recommended and each of the candidate similar users.

4. The method according to claim 3, characterized in that The determining of the recommended product according to the user similarity and the preference matrix includes: Determining, based on the user similarity and the preference matrix, the predicted interest of the to-be-recommended user for each of the candidate interest products in the preference matrix; Determine a recommended product based on the predicted interest level of each of the candidate products of interest.

5. The method according to claim 4, characterized in that The step of determining the predicted interest of the to-be-recommended user for each of the candidate interest products in the preference matrix according to the user similarity and the preference matrix includes: Determining target similar users of the user to be recommended according to the user similarity and the preset number of similar users; The predicted interest of the to-be-recommended user in each of the candidate interest products is determined according to the target similar users and the preference matrix.

6. The method according to claim 5, characterized in that The step of determining the predicted interest of the to-be-recommended user for each of the candidate interest products according to the target similar user and the preference matrix includes: Based on the target similar users and the preference matrix, the preference of each target similar user for each common product in the set of products visited by the user to be recommended, the average preference of the user to be recommended for each candidate product of interest, and the average preference of each target similar user for each candidate product of interest, the predicted visit rate of the user to be recommended for each candidate product of interest is determined.

7. An intelligent recommendation device, characterized in that: include: A user information acquisition module is used to obtain the access location of the user to be recommended and the product access rate of the candidate product; A target area determination module, used to determine a target area according to the access location, and obtain a product access rate of each user in the target area to each candidate product; A preference matrix determination module, used for clustering the product access rates of the to-be-recommended users to the candidate products and the product access rates of the users in the same region to the candidate products to obtain a preference matrix; A user similarity determination module, used to determine the user similarity between the to-be-recommended user and each candidate similar user in the preference matrix according to the preference matrix; The recommended product determination module is used to determine the recommended products according to the user similarity and the preference matrix.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the intelligent recommendation method as described in any one of claims 1-6 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the intelligent recommendation method as described in any one of claims 1 to 6 is implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, which, when executed by a processor, implements the intelligent recommendation method according to any one of claims 1 to 6.

Images