Recommendation apparatus, storage medium, and computer program product
By calculating the similarity of the target vector and similar user vectors for the new user, the recommendation system can accurately recommend the content of the new user's interest, and solve the problem of poor personalized recommendation effects in existing technology.
Patent Information
- Application Number
- CN202310886558.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-18
- Publication Date
- 2025-01-21
AI Technical Summary
In the prior art, personalized recommendations for new users cannot be made, resulting in poor recommendation results.
By determining the user's characteristic sequence and vector, calculate the target user vector of the target user, and match the similarity with the vector of other users to determine similar users, so as to recommend the target content.
It has achieved personalized recommendations for new users, significantly improved the recommendation effect, and ensures that the content that users are interested in are accurately recommended.
Smart Images
Figure CN119336974A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a recommendation device, a storage medium and a computer program product. Background Art
[0002] The CS (Cold Start) problem in the recall phase of the APP recommendation system (i.e., the phase of retrieving content that users are potentially interested in from a massive content library based on some of their characteristics) is becoming increasingly prominent. As new users (users) and new content (items) continue to arrive, if there is no efficient CS module in the recommendation system, new users and new content will not receive efficient personalized recommendations, which will greatly reduce user stickiness and content richness, leading to a decrease in the number of users and revenue.
[0003] In existing CS solutions, popular content, the latest content, or officially released content are usually recommended to new users when they arrive. Existing solutions have strong subjective factors and lack personalization by recalling some hot and fresh content for new users, resulting in poor recommendation results. Summary of the invention
[0004] The present invention provides a recommendation device, a storage medium and a computer program product, which are used to solve the defect in the prior art that personalized recommendations cannot be made for new users, resulting in poor recommendation effects.
[0005] In a first aspect, the present invention provides a recommendation device, comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program to implement the following steps:
[0006] Determine a user feature sequence according to the order in which the first users access the first content and the attribute characteristics of the first users, and determine a vector of each feature in the user feature sequence;
[0007] Determining a target vector of the target user according to the attribute characteristics of the target user and the vectors of the characteristics;
[0008] Determine the first vectors of the first users according to the attribute characteristics of the first users, and determine similar users according to the similarities between the target vector and the first vectors of the first users;
[0009] Target content is determined according to the similar users, and content is recommended to the target user according to the target content.
[0010] In one embodiment, the processor is further configured to filter the target content:
[0011] Acquire attribute feature information of multiple users who access first content to be selected; the first content to be selected is one of the contents accessed by the similar users;
[0012] Performing clustering processing on the attribute feature information of the plurality of users according to a preset clustering algorithm to generate user attribute features of the first to-be-selected content;
[0013] Performing similarity matching between the user attribute characteristics and the target user attribute characteristics;
[0014] When the similarity between the user attribute feature and the target user attribute feature is greater than or equal to a first similarity threshold, the first candidate content is determined to be the target content.
[0015] In one embodiment, the processor is further configured to filter the target content:
[0016] Obtaining user behaviors of the similar users;
[0017] According to a preset screening strategy, matching whether the user behavior reflects a strong behavior on the second candidate content;
[0018] When the user behavior reflects a strong behavior on the second content to be selected, the second content to be selected is determined as the target content.
[0019] In one embodiment, the processor is further configured to filter the first content by any of the following methods:
[0020] taking the content visited by the target user as the first content;
[0021] The content on which the target user has performed operations on the social platform is used as the first content.
[0022] In one embodiment, the processor is specifically configured to implement any of the following:
[0023] The first user whose similarity between the target vector and the first vector exceeds a similarity threshold is regarded as a similar user;
[0024] The first user whose target vector has the greatest similarity to the first vector is regarded as a similar user.
[0025] In one embodiment, the mandatory behavior includes at least one of the following:
[0026] The time spent browsing content exceeds the time threshold;
[0027] The number of times the content is accessed exceeds the threshold;
[0028] Like content;
[0029] Collection content;
[0030] Purchase content.
[0031] In one embodiment, the processor is specifically configured to implement any of the following:
[0032] Determine the cosine similarity between the target vector and the first vector of each first user;
[0033] Determine the distance similarity between the target vector and the first vector of each first user.
[0034] In one embodiment, the processor is specifically configured to:
[0035] Establish the index of the vector of each feature in the user feature sequence;
[0036] According to the attribute characteristics and index of the target user, determine the vector corresponding to each attribute characteristic of the target user;
[0037] The vectors corresponding to the attribute features of the target user are combined to obtain the target vector of the target user.
[0038] In a second aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed, implements the method steps executed by the processor as described in the first aspect.
[0039] In a third aspect, the present invention further provides a computer program product, comprising a computer program, which, when executed, implements the method steps executed by the processor as described in the first aspect.
[0040] The recommendation device, storage medium and computer program product provided by the embodiments of the present invention can achieve the purpose of accurately recommending content of interest to the user to be recommended, thereby achieving the purpose of making personalized recommendations for new users and significantly improving the recommendation effect, by performing similarity expansion between the vector of the target user and the vectors of multiple users who access the first content, to determine multiple similar users with a high similarity to the target user, and recommending content to the target user based on the similar users. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0042] Figure 1 is a flowchart of the steps of the method performed by the recommendation device provided in an embodiment of the present invention;
[0043] Figure 2 is a schematic diagram of the structure of a recommendation system using the recommendation device provided by an embodiment of the present invention;
[0044] Figure 3 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0045] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0046] Figure 1 is a flow chart of the steps of the method performed by the recommendation device provided in an embodiment of the present invention. Figure 1 The recommendation device provided in the embodiment of the present invention may include: a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program to implement the following steps:
[0047] Step 110: determining a user feature sequence according to the order in which the first users access the first content and the attribute characteristics of the first users, and determining a vector of each feature in the user feature sequence;
[0048] Step 120: Determine a target vector of the target user based on the attribute characteristics of the target user and the vectors of each characteristic;
[0049] Step 130: determining a first vector of each first user according to the attribute characteristics of each first user, and determining similar users according to the similarity between the target vector and the first vector of each first user;
[0050] Step 140: Determine target content based on similar users, and recommend content to target users based on the target content.
[0051] It should be noted that the recommendation device provided in the embodiment of the present invention may specifically be a computer device, such as a mobile phone, a tablet computer, a laptop computer, a PDA, etc.
[0052] Before executing step 110, the recommendation device may use a computing engine such as Spark to clean the attribute feature data of each first user who accesses the first content from a data warehouse such as Hive. Alternatively, the recommendation device may directly call the attribute feature data of each first user who accesses the first content stored in the database through an interface.
[0053] The embodiment of the present invention does not specifically limit the specific method of obtaining the attribute characteristics of the first user.
[0054] In step 110 , the recommendation device may generate a user feature sequence by using the order in which the first users access the first content and the attribute features of the first users.
[0055] The first user may be any user. The first content may be any content, such as various commodities, books, videos, music, etc.
[0056] For example, the recommendation device can use a computing engine such as Spark to generate a user feature sequence by chronologically accessing content A by user a, user b, user c, ..., user n, as well as the attribute characteristics of each user an, such as at least one of id, gender, age, population label, occupation, height, weight, lifestyle, and behavior.
[0057] Among them, the population labels can be obtained by dividing the population according to a certain classification algorithm. For example, according to the income classification algorithm, the population labels such as "rich", "high-yield", "middle class", and "wage earner" can be obtained; according to the height-to-weight ratio classification algorithm, the population labels such as "obese", "slightly fat", "symmetrical", and "thin" can be obtained.
[0058] Lifestyle can be summarized into user labels based on the user's food, clothing, housing and transportation data, such as "homebody", "homegirl", "sunshine guy", "fitness expert", "outdoor expert", "internet addict teenager", etc.
[0059] After determining the user feature sequence, the recommendation device can determine the vector of each feature in the user feature sequence. For example, the recommendation device can input the user feature sequence into an algorithm implementation tool such as Tensorflow, process the user feature sequence using a model with a text-to-vector function such as Word2vec, obtain the vector of each feature in the user feature sequence, and cache the processing result in a database such as Redis.
[0060] It can be understood that since the user feature sequence can be obtained based on the attribute characteristics of different users accessing the same content, and the correlation and similarity between different users accessing the same content are relatively strong, the use of the user feature sequence in the embodiment of the present application can help match similar users, thereby improving the recommendation efficiency and recommendation effect.
[0061] In step 120 , the recommendation device may determine a target vector of the target user based on the attribute characteristics of the target user and the vectors of each feature in the user feature sequence obtained in step 110 .
[0062] For example, the recommendation device can first use tools such as Flink to read the attribute characteristics of the target user in the Kafka system, and match the vectors of each feature corresponding to the attribute characteristics of the target user from the user feature sequence according to the attribute characteristics of the target user, and generate the target vector of the target user according to the corresponding vectors of each feature. Subsequently, the recommendation device can cache the target vector in a database such as Redis.
[0063] The target user may be the user to whom the content is to be recommended.
[0064] In step 130 , the recommendation device may determine the first vector of each first user according to the attribute characteristics of each first user, and determine one or more similar users according to the similarity between the target vector and the first vector of each first user.
[0065] For example, the recommendation device may input the attribute features of the first user into an algorithm implementation tool such as Tensorflow, and use a model with a text-to-quantity function such as DSSM to process the attribute features of the first user, thereby obtaining the first vector of the first user. Alternatively, the recommendation device may also match the vectors of each feature corresponding to the attribute features of the first user from the user feature sequence, and generate the first vector of the first user based on the corresponding vectors of each feature.
[0066] Similar users are users who have a high degree of similarity with the target user. It should be noted that high similarity can refer to high similarity in the attribute characteristics of the users, such as people with the same label are all "highly productive", people with the same height and weight are all "well-proportioned", people with the same lifestyle are all "homebodies", etc.
[0067] The recommendation device may read the target vector and each first vector stored in a database such as Redis, then calculate the similarity between the target vector and each first vector, and determine one or more similar users from each first user according to the similarity determination result.
[0068] In step 140 , the recommendation device may determine target content according to similar users, and recommend content to the target user according to the target content.
[0069] For example, the recommendation device may first determine the target content that the similar users are interested in, and recommend the target content of interest to the target user.
[0070] The target content may be content in which similar users have strong behaviors, where strong behaviors may refer to behaviors that show interest in the content, such as collection behaviors, purchase behaviors, etc.
[0071] For example, when the similar user is user a, and content A and content B are contents that user a embodies strong behavior, the recommendation device will recommend content A and content B to the target user; if the similar users are user a, user b and user c, and the contents that user a embodies strong behavior are content A and content B, the contents that user b embodies strong behavior are content A and content C, and the content that user c embodies strong behavior is content D, the recommendation device will recommend content A, content B, content C and content D to the target user.
[0072] The target content may also be content that satisfies the following condition: the attribute characteristics of the user accessing the target content are similar to the attribute characteristics of the target user.
[0073] It is understandable that the attribute characteristics of users accessing the target content can reflect the regularity of users who are interested in the target content. Furthermore, when the attribute characteristics of users accessing a certain content are similar to the attribute characteristics of the target user, the content is recommended to the target user as the target content, which can ensure that the target user can be recommended the content that he or she is interested in, thereby improving the recommendation effect.
[0074] The recommendation device provided by the embodiment of the present invention determines similar users with high similarity to the target user by performing similarity expansion between the vector of the target user and the vectors of multiple users who access the first content, and recommends content to the target user based on the similar users. This can achieve the purpose of accurately recommending content of interest to the user to be recommended, thereby achieving the purpose of making personalized recommendations for new users and significantly improving the recommendation effect.
[0075] In one embodiment, the processor of the recommendation device may also be used to filter target content:
[0076] Acquiring attribute feature information of multiple users who access the first content to be selected; the first content to be selected is one of the contents accessed by similar users;
[0077] Clustering the attribute feature information of multiple users according to a preset clustering algorithm to generate user attribute features of the first candidate content;
[0078] Perform similarity matching between the user's attribute characteristics and the target user's attribute characteristics;
[0079] When the similarity between the attribute characteristics of the user and the attribute characteristics of the target user is greater than or equal to a first similarity threshold, the first candidate content is determined to be the target content.
[0080] The recommendation device can obtain attribute feature information of users who have historically visited the first content to be selected, or obtain attribute feature information of users who have visited the first content to be selected within a period of time (for example, within a week or a month, etc.), and cluster the acquired attribute feature information of each user to obtain user attribute features of the first content to be selected.
[0081] For example, for the first candidate content A, the recommendation device may obtain attribute feature information of each user who accesses content A within a week, and cluster the attribute feature information of each user through a preset clustering algorithm, such as a K-means clustering algorithm, a GMM Gaussian mixture model clustering algorithm, etc., to obtain user attribute features of the first candidate content. The user attribute features may be the clustering results of various attribute features such as the user's gender, age, the population label, occupation, and lifestyle.
[0082] Next, the recommendation device may calculate the similarity between the user attribute characteristics of the first candidate content and the attribute characteristics of the target user.
[0083] For example, the recommendation device may calculate the similarity between the user attribute characteristics of the first candidate content and the attribute characteristics of the target user to obtain a similarity value M. If M is greater than or equal to a first similarity threshold M1, the first candidate content A may be determined as the target content.
[0084] If M is less than the similarity threshold M1, it can be determined that the first candidate content A is not the target content.
[0085] It is understandable that by clustering the attribute feature information of multiple users who access the first candidate content, the regularity of users who are interested in the first candidate content can be determined. Furthermore, the similarity between the user attribute features of the first candidate content obtained by clustering and the attribute features of the target user can reflect the possibility that the first candidate content is the content of interest to the target user.
[0086] Therefore, by recommending the first candidate content to the target user when the similarity between the user attribute characteristics of the first candidate content and the attribute characteristics of the target user is greater than or equal to the first similarity threshold, it can be ensured that the target user can be recommended the content that he or she is interested in, thereby improving the recommendation effect.
[0087] In one embodiment, the processor of the recommendation device may also be used to filter target content:
[0088] Get user behavior of similar users;
[0089] According to the preset screening strategy, whether the matching user behavior reflects a strong behavior for the second candidate content;
[0090] When the user behavior reflects a strong behavior on the second candidate content, the second candidate content is determined as the target content.
[0091] The recommendation device may obtain the historical user behavior of each similar user, or obtain the user behavior of each similar user within a period of time (e.g., within a week or a month, etc.). User behavior refers to the operation behavior of the user on the content, such as liking, collecting, purchasing, etc. the content.
[0092] The recommendation device may determine whether the similar users have shown strong behavior towards the second candidate content according to a preset screening strategy.
[0093] The preset screening strategy may include determining whether the similar user exhibits a strong behavior on the second candidate content according to whether the operation data of the similar user on the second candidate content reaches a threshold corresponding to the operation.
[0094] For example, when the time that a similar user browses a second selected content exceeds a time threshold, it can be considered that the similar user has shown strong behavior towards the second selected content. Or, when the number of times a similar user visits a second selected content exceeds a number threshold, it can be considered that the similar user has shown strong behavior towards the second selected content.
[0095] The preset screening strategy may also include determining whether the similar user has a strong behavior on the second to-be-selected content according to whether the operation of the similar user on the second to-be-selected content is a specific operation.
[0096] For example, when a similar user has liked a second selected content, it can be considered that the similar user has shown strong behavior towards the second selected content. Or, when a similar user has collected a second selected content, it can be considered that the similar user has shown strong behavior towards the second selected content. Or, when a similar user has purchased a second selected content, it can be considered that the similar user has shown strong behavior towards the second selected content.
[0097] The recommendation device provided in the embodiment of the present application determines whether the second candidate content is the target content by judging whether similar users have shown strong behavior on the second candidate content. It can accurately determine the content that similar users are interested in, and then by recommending the target content to the target user, it can ensure that the target user is highly likely to be recommended the content that he or she is interested in, thereby ensuring the recommendation effect.
[0098] In one embodiment, the processor of the recommendation device may also be configured to filter the first content by any of the following methods:
[0099] Use the content that the target user has visited as the first content;
[0100] Use the content that the target users have operated on the social platform as the first content.
[0101] It should be noted that the user feature sequence is obtained based on different users accessing the same content, and the correlation and similarity between users accessing the same content are relatively strong.
[0102] Therefore, when the target user has visited a certain content, it means that the target user is likely to be interested in the content. Therefore, using the content as the first content can not only ensure that the correlation and similarity between users who visit the same content are relatively strong, but also ensure that the correlation and similarity between each user and the content of interest of the target user are relatively strong, thereby further ensuring the recommendation effect.
[0103] Similarly, when the target user has operated a certain content on a social platform, it can also indicate that the target user is more likely to be interested in the content. Therefore, using the content as the first content can further ensure the recommendation effect.
[0104] In one embodiment, the processor of the recommendation device may be specifically configured to implement any of the following:
[0105] 1. The first user whose similarity between the target vector and the first vector exceeds the similarity threshold is regarded as a similar user;
[0106] The similarity threshold may be, for example, 80%, 90%, etc., and its specific value may be adjusted according to actual conditions, which is not specifically limited in the embodiment of the present invention.
[0107] For example, the recommendation device calculates the similarity between the target vector A of the target user a and the first vector B of the first user b, the first vector C of the first user c, ..., the first vector N of the first user n, and determines that the similarity between the target vector A and the first vector B, and the similarity between the target vector A and the first vector C exceed the similarity threshold. In this case, the recommendation device will regard the first user b and the first user c as similar users of the target user a.
[0108] 2. The first user whose target vector has the greatest similarity to the first vector is regarded as a similar user.
[0109] For example, the recommendation device calculates the similarity between the target vector A of the target user a and the first vector B of the first user b, the first vector C of the first user c, ..., and the first vector N of the first user n, and determines that the similarity between the target vector A and the first vector N is the greatest. In this case, the recommendation device will select the first user n as a similar user of the target user a.
[0110] The recommendation device provided by the embodiment of the present invention can determine multiple similar users in multiple ways, and can adapt to the content recommendation needs in various scenarios, so as to ensure that the content recommended to the recommended user is more personalized and the recommendation effect is guaranteed.
[0111] In one embodiment, the mandatory behavior may include at least one of the following:
[0112] 1. The duration of browsing content exceeds the duration threshold;
[0113] The duration threshold may be, for example, 30 seconds, 60 seconds, etc., and its specific value may be adjusted according to actual conditions, and the embodiment of the present invention does not specifically limit this.
[0114] For example, when the browsing time of the second candidate content by the similar user exceeds a time threshold, such as 30 seconds, the recommendation device may determine that the similar user has shown strong behavior on the second candidate content and recommend the second candidate content to the target user.
[0115] 2. The number of times the content is accessed exceeds the threshold;
[0116] The number threshold may be, for example, 2 times, 5 times, etc., and its specific value may be adjusted according to actual conditions, and the embodiment of the present invention does not specifically limit this.
[0117] For example, when the number of visits to the second candidate content by similar users exceeds a threshold, such as 2 times, the recommendation device may determine that the similar users have shown strong behavior towards the second candidate content and recommend the second candidate content to the target user.
[0118] 3. Like the content;
[0119] When the similar user has liked the second candidate content, the recommendation device may determine that the similar user has shown a strong behavior towards the second candidate content, and recommend the second candidate content to the target user.
[0120] 4. Collect content;
[0121] When the similar user has collected the second content to be selected, the recommendation device may determine that the similar user has shown a strong behavior towards the second content to be selected, and recommend the second content to be selected to the target user.
[0122] 5. Purchase content.
[0123] When the similar user has purchased the second content to be selected, the recommendation device may determine that the similar user has shown a strong behavior towards the second content to be selected, and recommend the second content to be selected to the target user.
[0124] The recommendation device provided in the embodiment of the present application determines whether the user has shown strong behavior towards the content by judging whether the time the user has browsed the content exceeds a threshold, whether the number of times the user has visited the content exceeds a threshold, whether the user has liked, collected, or purchased the content, etc., and can accurately determine whether the user is interested in the content, so as to specifically recommend the content of interest to the target user, thereby ensuring the recommendation effect.
[0125] In one embodiment, the processor of the recommendation device may be specifically configured to implement any of the following:
[0126] 1. Determine the cosine similarity between the target vector and the first vector of each first user;
[0127] After determining the target vector of the target user and the first vectors of each first user, the recommendation device may determine the cosine value of the angle between the target user and each first vector, and then determine the similarity between the target vector and each first vector according to each cosine value.
[0128] 2. Determine the distance similarity between the target vector and the first vector of each first user;
[0129] After determining the target vector of the target user and the first vectors of each first user, the recommendation device can determine the distance similarity between the target vector and each first vector, such as Minkowski distance similarity, Manhattan distance similarity, Euclidean distance similarity, Chebyshev distance similarity, etc.
[0130] The recommendation device provided in the embodiment of the present invention determines the similarity between the target user and each first user in a variety of ways, and can adapt to the similarity calculation requirements in various scenarios, thereby ensuring accurate content recommendation to the target user and ensuring the recommendation effect.
[0131] In one embodiment, the processor of the recommendation device may be specifically configured to:
[0132] Establish the index of the vector of each feature in the user feature sequence;
[0133] According to the attribute characteristics and index of the target user, determine the vector corresponding to each attribute characteristic of the target user;
[0134] The vectors corresponding to the attribute features of the target content are combined to obtain the target vector of the target user.
[0135] After determining the vector of each feature of the user feature sequence, the recommendation device may create an index of the feature and the vector, and cache the index in a database such as Redis.
[0136] Afterwards, the recommendation device can read the index in a database such as Redis, and match the vectors corresponding to each attribute feature from the index according to the attribute features of the target user, and combine the vectors corresponding to each attribute feature to obtain the target vector of the target user.
[0137] The specific method of combination may be vector splicing, such as horizontal splicing of vectors.
[0138] The recommendation device provided by the embodiment of the present invention can ensure the accuracy of the target vector by establishing an index of the vector of each feature in the user feature sequence and obtaining the target vector of the target user according to the attribute characteristics of the target user and the index, thereby improving the accuracy of determining similar users.
[0139] The following uses an example of a recommended device provided by an embodiment of the present invention as an example to further illustrate the technical solution of the present invention:
[0140] This example may be a recommendation system including the recommendation device provided in the embodiment of the present application. The structure of the recommendation system is as follows: Figure 2 As shown, it includes: data module, Word2vec module, DSSM module, and expansion module.
[0141] The data module is used to clean the historical data of each first user, obtain the attribute characteristics of each first user, and generate a user characteristic sequence.
[0142] The DSSM module is used to train the DSSM model based on the attribute feature data of historical users, and input the attribute features of each first user into the trained DSSM model to obtain each first vector.
[0143] The Word2vec module is used to train the Word2vec model based on the user feature sequence and input the attribute features of the target user into the trained Word2vec model to obtain the target vector of the target user;
[0144] The expansion module is used to calculate the similarity between the target vector and each first vector, thereby determining one or more similar users, and recommending content to the target user based on the similar users.
[0145] Figure 3 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 3As shown, the electronic device may include: a processor 310, a communication interface 320, a memory 330 and a communication bus 340, wherein the processor 310, the communication interface 320 and the memory 330 communicate with each other through the communication bus 340. The processor 310 may call the logic instructions in the memory 330 to execute the method steps executed by the processor of each embodiment above, for example, including:
[0146] Determine a user feature sequence according to the order in which the first users access the first content and the attribute characteristics of the first users, and determine a vector of each feature in the user feature sequence;
[0147] Determine the target vector of the target user according to the attribute characteristics of the target user and the vector of each characteristic;
[0148] Determine a first vector of each first user according to the attribute characteristics of each first user, and determine similar users according to the similarity between the target vector and the first vector of each first user;
[0149] Determine target content based on similar users, and recommend content to target users based on the target content.
[0150] In addition, the logic instructions in the above-mentioned memory 330 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.
[0151] On the other hand, the present invention further provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed, the computer can execute the method steps executed by the processor of each of the above embodiments, for example, including:
[0152] Determine a user feature sequence according to the order in which the first users access the first content and the attribute characteristics of the first users, and determine a vector of each feature in the user feature sequence;
[0153] Determine the target vector of the target user according to the attribute characteristics of the target user and the vector of each characteristic;
[0154] Determine a first vector of each first user according to the attribute characteristics of each first user, and determine similar users according to the similarity between the target vector and the first vector of each first user;
[0155] Determine target content based on similar users, and recommend content to target users based on the target content.
[0156] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed, implements the method steps executed by the processor of each of the above embodiments, for example, including:
[0157] Determine a user feature sequence according to the order in which the first users access the first content and the attribute characteristics of the first users, and determine a vector of each feature in the user feature sequence;
[0158] Determine the target vector of the target user according to the attribute characteristics of the target user and the vector of each characteristic;
[0159] Determine a first vector of each first user according to the attribute characteristics of each first user, and determine similar users according to the similarity between the target vector and the first vector of each first user;
[0160] Determine target content based on similar users, and recommend content to target users based on the target content.
[0161] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.
[0162] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is 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 ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.
[0163] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A recommendation device, characterized in that: include: A memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program to implement the following steps: Determine a user feature sequence according to the order in which the first users access the first content and the attribute characteristics of the first users, and determine a vector of each feature in the user feature sequence; Determining a target vector of the target user according to attribute characteristics of the target user and vectors of the characteristics; Determine the first vectors of the first users according to the attribute characteristics of the first users, and determine similar users according to the similarities between the target vector and the first vectors of the first users; Target content is determined according to the similar users, and content is recommended to the target user according to the target content.
2. The recommendation device according to claim 1, characterized in that: The processor is further configured to filter the target content: Acquire attribute feature information of multiple users who access first content to be selected; the first content to be selected is one of the contents accessed by the similar users; Performing clustering processing on the attribute feature information of the plurality of users according to a preset clustering algorithm to generate user attribute features of the first to-be-selected content; Performing similarity matching between the user attribute characteristics and the target user attribute characteristics; When the similarity between the user attribute feature and the target user attribute feature is greater than or equal to a first similarity threshold, the first candidate content is determined to be the target content.
3. The recommendation device according to claim 1, characterized in that: The processor is further configured to filter the target content: Obtaining user behaviors of the similar users; According to a preset screening strategy, matching whether the user behavior reflects a strong behavior on the second candidate content; When the user behavior reflects a strong behavior on the second content to be selected, the second content to be selected is determined as the target content.
4. The recommendation device according to claim 1, characterized in that: The processor is further configured to filter the first content by any of the following methods: taking the content visited by the target user as the first content; The content on which the target user has performed operations on the social platform is used as the first content.
5. The recommendation device according to claim 1, characterized in that: The processor is specifically configured to implement any of the following: The first user whose similarity between the target vector and the first vector exceeds a second similarity threshold is used as the similar user; The first user whose similarity between the target vector and the first vector is the greatest is used as the similar user.
6. The recommendation device according to claim 3, characterized in that: The mandatory behavior includes at least one of the following: The time spent browsing content exceeds the time threshold; The number of times the content is accessed exceeds the threshold; Like content; Collection content; Purchase content.
7. The recommendation device according to claim 1, characterized in that: The processor is specifically configured to implement any of the following: Determine the cosine similarity between the target vector and the first vector of each first user; Determine the distance similarity between the target vector and the first vector of each first user.
8. The recommendation device according to any one of claims 1 to 7, characterized in that: The processor is specifically used for: Establishing an index of a vector of each feature in the user feature sequence; Determine, according to the attribute characteristics of the target user and the index, a vector corresponding to each attribute characteristic of the target user; The vectors corresponding to the attribute features of the target user are combined to obtain a target vector of the target user.
9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the method steps executed by the processor according to any one of claims 1 to 8 are implemented.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed, the method steps executed by the processor according to any one of claims 1 to 8 are implemented.