Information recommendation method, device, equipment and medium
By analyzing user behavior data, generating reference numbers and target classifications, exploratory recalls and coordinated recommendations are achieved, and the problems of narrowing the coverage of recommended content and independent recall strategies in the existing technology are solved, and the accuracy and coverage of information recall are improved.
Patent Information
- Application Number
- CN201910443788.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2019-05-27
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2040-02-11
AI Technical Summary
The existing recommendation system recalls based on user portraits, which cannot predict the user's future behavior, resulting in the gradual reduction of the coverage of recommendation content, and the various recall strategies are independent and uncoordinated, so it is impossible to take global considerations to obtain objective recommendations.
By obtaining user behavior data, generating references for each initial classification, extracting target classifications, and obtaining information from target information sets, to achieve exploratory recall and coordinated recommendations, ensuring that users see untouched information and optimize recall results.
It improves the accuracy and coverage of information recalls, broadens the diversity and richness of recommended information, and avoids the problem of narrowing the coverage of recommended content.
Smart Images

Figure CN110321478B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of information recommendation, and in particular to an information recommendation method, device, equipment and medium. Background Art
[0002] Existing recommendation systems usually use various recall strategies to obtain a recommendation list based on user portraits, and present recommended content to users based on the recommendation list. However, this technical solution has the following problems:
[0003] (1) Recall strategies are all based on user portraits, and user portraits can only describe the user's historical behavior. Therefore, the recall results of the recall strategy only refer to the user's historical behavior, and cannot predict the user's future behavior. For example, each recalled article is related to the user's historical click behavior. The related articles that the user has read before will continue to appear, and the user will not have the opportunity to read other articles. The recall strategy cannot recall content that the user has not been exposed to, resulting in an increasingly narrow coverage of recommended content.
[0004] (2) Each recall strategy in the recommendation system operates independently, and there is no overall coordination of various recall strategies. It is also impossible to make a global consideration based on the recall results corresponding to various recall strategies to obtain objective recommendation content. Summary of the invention
[0005] In order to solve the technical problem in the prior art that only user portraits are considered for recommendation, resulting in increasingly narrow coverage of recommended content, the embodiments of the present invention provide an information recommendation method, device, equipment and medium.
[0006] In one aspect, the present invention provides an information recommendation method, the method comprising:
[0007] Acquire user behavior data, where the user behavior data is used to characterize the user's interest in various initial categories;
[0008] Generate a reference number for each initial classification, and extract a target classification from the initial classification according to the reference number; the generation of the reference number conforms to a probability distribution, and the probability distribution is related to the user behavior data of the initial classification corresponding to the reference number;
[0009] Obtaining the corresponding target information in each target classification to obtain a target information set;
[0010] Information for recommendation is extracted from the target information set.
[0011] In another aspect, the present invention provides an information recommendation device, the device comprising:
[0012] A user behavior data or acquisition module, used to acquire user behavior data, wherein the user behavior data is used to characterize the user's interest in various initial categories;
[0013] A user behavior data analysis module, used to generate a reference number for each initial classification, and extract a target classification from the initial classification according to the reference number; the generation of the reference number conforms to a probability distribution, and the probability distribution is related to the user behavior data of the initial classification corresponding to it;
[0014] A target information set acquisition module is used to acquire the target information corresponding to each target classification to obtain a target information set;
[0015] The recommendation information extraction module is used to extract information for recommendation from the target information set.
[0016] On the other hand, the present invention provides a device, characterized in that the device includes a processor and a memory, the memory storing at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set are loaded and executed by the processor to implement an information recommendation method.
[0017] On the other hand, the present invention provides a computer storage medium, characterized in that the storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded by a processor and executes an information recommendation method.
[0018] The present invention provides an information recommendation method, device, equipment and medium. In the present invention, a target classification that is compatible with user portraits and can explore user interests is obtained through analysis of user behavior data, thereby realizing an exploratory recall, and each information in the target classification is extracted again, thereby realizing a comprehensive recommendation. Furthermore, the exploratory recall provided by the present invention enables users to have the opportunity to see information under unexposed classifications, realizes overall optimization modeling of various recall results, and improves the accuracy and coverage of information recall. By continuously exploring user interests, the diversity and richness of user recommendation information can be broadened, avoiding the problem of increasingly narrow coverage of recommended information. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only 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.
[0020] Figure 1 It is a schematic diagram of an implementation environment provided by the present invention;
[0021] Figure 2 It is a flow chart of an information recommendation method provided by the present invention;
[0022] Figure 3 This is a schematic diagram of initial classification in a news recommendation scenario provided by the present invention;
[0023] Figure 4 It is a schematic diagram of initial classification in a book recommendation scenario provided by the present invention;
[0024] Figure 5 It is a schematic diagram of initial classification in a video recommendation scenario provided by the present invention;
[0025] Figure 6 It is a flow chart of a method for smoothly changing user behavior data provided by the present invention;
[0026] Figure 7 is a flow chart of generating reference numbers for each initial classification provided by the present invention;
[0027] Figure 8 The present invention provides a flowchart for extracting a target classification from the initial classification according to the reference number;
[0028] Fig. 9 The present invention provides a flowchart for obtaining target information corresponding to each target classification to obtain a target information set;
[0029] Fig.10 is a schematic diagram of an information recommendation method provided by the present invention;
[0030] Fig.11 It is a schematic diagram of an embodiment of the present invention provided by the present invention in a news scenario at a user terminal;
[0031] Fig.12 It is a block diagram of an information recommendation device provided by the present invention;
[0032] Fig.13 It is a block diagram of a user behavior data smooth change module provided by the present invention;
[0033] Fig.14 It is a schematic diagram of the hardware structure of a device provided by the present invention for implementing the method provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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 making creative work are within the scope of protection of the present invention.
[0035] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention 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 invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes 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.
[0036] In order to make the purpose, technical solution and advantages disclosed in the embodiments of the present invention more clearly understood, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not used to limit the embodiments of the present invention.
[0037] Figure 1 is a schematic diagram of an implementation environment provided by an embodiment of the present invention, see Figure 1 The implementation environment includes: a recommendation server 01 and at least one recommendation terminal 02, wherein the recommendation server 01 is in communication connection with the recommendation terminal 02.
[0038] The recommendation terminal 02 can communicate with the recommendation server 01 based on the browser / server mode (B / S) or the client / server mode (C / S). The recommendation terminal 02 can include physical devices such as smart phones, tablet computers, laptops, digital assistants, smart wearable devices, and vehicle terminals, and can also include software running in physical devices, such as applications. For example, the recommendation terminal can run news recommendation software, video recommendation software, microblog recommendation software, and shopping recommendation software.
[0039] The recommendation server 01 is used to obtain the user portrait of the customer corresponding to each recommendation terminal 02, obtain the recommended content based on the user portrait, and push the recommended content to each recommendation terminal 02. The recommendation server 01 may include an independently operated server, or a distributed server, or a server cluster composed of multiple servers.
[0040] Please refer to Figure 2 , which shows an information recommendation method provided by an embodiment of the present invention. The method can be implemented by the recommendation server in the above implementation environment, and the method includes:
[0041] S101. Obtain user behavior data, where the user behavior data is used to characterize the user's interest in various initial categories.
[0042] Specifically, the initial classification may be different in different scenarios. The initial classification may be the coarsest classification in a specific scenario.
[0043] like Figure 3 As shown in Figure 1, in the news recommendation scenario, the information classification of news recommendation is the initial classification. The initial classification of news is the first-level classification of the news page. Figure 3 It includes society, entertainment, news, sports, life circle, Greater Suzhou, finance, and people's livelihood.
[0044] like Figure 4 As shown, in the book recommendation scenario, the classification information of the book is the initial classification. Figure 4 It includes romance, martial arts, mythology, science fiction, tomb raiding, and mystery.
[0045] like Figure 5 As shown, in the video recommendation scenario, the classification information of the video is the initial classification. Figure 5 Includes homemade, costume, romance, idol, family, youth, urban, comedy, and spy war.
[0046] In a feasible embodiment, the user behavior data can be represented by the user's unselected value data and the user's selected value data. Specifically, the user's unselected value data is the difference between the data exposed to the user and the user's selected value data. User behavior data is a user portrait, which represents the user's interest orientation for each initial classification. User portrait is a description data of multi-dimensional user behavior habit attributes based on the user's historical behavior.
[0047] The user behavior data can be captured in real time through the Spark platform and stored in real time based on Redis.
[0048] The Spark platform is a fast and versatile computing engine designed for large-scale data processing. Spark is an open source cluster computing environment that enables in-memory distributed datasets. In addition to providing interactive queries, it can also optimize iterative workloads.
[0049] Redis is a key-value storage system. Redis supports master-slave synchronization. Data can be synchronized from a master server to any number of slave servers, which can be a master server associated with other slave servers. This allows Redis to perform single-layer tree replication. Data can be written intentionally or unintentionally when saving to disk. Since the publish / subscribe mechanism is fully implemented, when the slave database synchronizes the tree anywhere, it can subscribe to a channel and receive the complete message publishing record of the master server. Synchronization is very helpful for scalability of read operations and data redundancy.
[0050] Specifically, users can select data through various actions or gestures such as clicking, sliding, circling, etc.
[0051] In a feasible implementation, the user behavior data of a certain user in a certain time period for a certain initial classification can be represented by a user behavior data pair, that is, a user behavior data pair (a: selected value, b: unselected value), and the corresponding user behavior data of a certain user in multiple time periods (or a time interval consisting of multiple consecutive time periods) for a certain initial classification can be represented by a user behavior data pair sequence. Therefore, the user behavior of each user in each initial classification can also be represented by a user behavior data pair sequence.
[0052] For example, taking a day as a time period, the user behavior data of a certain user for a certain initial classification within 30 consecutive days can be represented by a user behavior data pair sequence including 30 user behavior data pairs.
[0053] In a preferred embodiment, for the initial classification that has been exposed to the user, its user behavior data pair records the real user behavior, and for the initial classification that has not been exposed to the user, its user behavior data pair records the virtual user behavior, and the virtual user behavior is represented by a data with a smaller absolute value. For example, for the initial classification <Entertainment> that has not been exposed to the user, its user behavior data pair can be (0, 0.001), 0 represents the selected value, 0.001 represents the selected value, and accordingly, its exposure number is 0+0.001=0.001.
[0054] Since users have obvious subjective tendencies in their interest orientations for each initial classification, the differences in the values of the user behavior data pairs under different classifications are relatively large, and the value fluctuations of different user behavior data pairs in a user behavior data pair sequence are also relatively large. Each selected or unselected behavior of a user will cause a change in the data in a certain user behavior data pair, which may further aggravate the differences between each user behavior data pair. In order to obtain relatively smooth user behavior data, in the user behavior data generation stage, an embodiment of the present invention discloses a method for smoothly changing user behavior data, such as Figure 6 As shown, including:
[0055] S1. In response to a user instruction, determine a target user behavior data pair pointed to by the user instruction.
[0056] Specifically, the user instruction may be an instruction for selecting or not selecting one or some information in a certain target initial classification. The user behavior data pair corresponding to the target initial classification is the target user behavior data pair.
[0057] S2. Obtain the sum of the selected value and the non-selected value of the target user behavior data pair.
[0058] S3. According to the user instruction, determine the target data to be changed in the target user behavior data pair.
[0059] Specifically, the target user behavior data pair is composed of (selected value, unselected value). If the user instruction is a selected instruction, such as clicking on an information chart in the target initial classification, the target data to be changed is the selected value in the target user behavior data pair; if the user instruction is a unselected instruction, such as directly closing the page corresponding to all information in the target initial classification, the target data to be changed is the unselected value in the target user behavior data pair.
[0060] S4. If the sum value is not greater than the preset threshold, the data to be changed is incremented by one.
[0061] S5. If the sum value is greater than the preset threshold, the data to be changed is incremented by one and then multiplied by a smoothing coefficient related to the preset threshold.
[0062] Specifically, taking the preset threshold value of 30 as an example, in the target user behavior data pair, a represents the selected value, and b represents the unselected value. When (a+b)<=30, the selected instruction a+1 is responded to, and the unselected instruction b+1 is responded to; when (a+b)>30, the selected instruction (a+1)*(30 / 31) is responded to, and the unselected instruction (b+1)*(30 / 31) is responded to, where 30 / 31 is the smoothing coefficient.
[0063] This smooth change can achieve the following effect: as user behavior data accumulates, the sum of the selected value and the non-selected value is no greater than a preset threshold and can reflect the user's behavior orientation, where the preset threshold can be set according to the effect of the time scenario.
[0064] S103. Generate a reference number for each initial classification, and extract a target classification from the initial classification according to the reference number; the generation of the reference number conforms to a probability distribution, and the probability distribution is related to the user behavior data of the initial classification corresponding to it.
[0065] Specifically, the user behavior data may be a sequence of user behavior data pairs corresponding to each initial classification within a preset period. If the preset period is 30 days, the sequence of user behavior data pairs is obtained based on the user's behavior within the past 30 days.
[0066] Specifically, the target classification can be obtained based on the user behavior data corresponding to each of the initial classifications and the data processing results that conform to statistical laws. The target classification is a classification that can be used to comprehensively explore the user's interests, which may include classifications that the user has been interested in in the past, and classifications that the user has rarely had the opportunity to contact in the past.
[0067] Specifically, in a feasible embodiment, the reference number of each initial classification is generated, such as Figure 7 As shown, including:
[0068] S1031. Obtain a target user behavior data pair corresponding to each initial classification according to the user behavior data, wherein the target user behavior data pair includes a selected value and a non-selected value.
[0069] Specifically, the target user behavior data pair is in a one-to-one correspondence with the initial classification, and the target user behavior data pair includes a selected value and a non-selected value for the initial classification.
[0070] In a feasible implementation, the selected value in the target user behavior data pair corresponding to the initial classification may be the sum of the selected values in each user behavior data pair in the user behavior data pair sequence corresponding to the initial classification; correspondingly, the unselected value in the target user behavior data pair corresponding to the initial classification may be the sum of the unselected values in each user behavior data pair in the user behavior data pair sequence corresponding to the initial classification.
[0071] In another feasible implementation, the selected values in the target user behavior data pairs corresponding to the initial classification can be extracted from the selected values of each user behavior data pair in the user behavior data pair sequence corresponding to the initial classification based on a preset sampling method; correspondingly, the unselected values in the target user behavior data pairs corresponding to the initial classification can be extracted from the unselected values of each user behavior data pair in the user behavior data pair sequence corresponding to the initial classification based on the preset sampling method.
[0072] S1033. For each initial classification, a reference number is generated based on its corresponding probability distribution, and the expected value of the probability distribution is related to the result of dividing the selected value in the target user behavior pair corresponding to the initial classification by the sum of the selected value and the non-selected value.
[0073] Specifically, the expected value may be equal to, proportional to, or positively correlated to the result of dividing the selected value by the sum of the selected value and the non-selected value in the target user behavior pair corresponding to the initial classification.
[0074] Specifically, the probability distribution may be a beta probability distribution, a gamma distribution or others, and the embodiment of the present invention does not limit the specific probability distribution form.
[0075] Beta distribution is a density function that is a conjugate prior distribution of Bernoulli distribution and binomial distribution, and has important applications in machine learning and mathematical statistics. In probability theory, Beta distribution, also known as B distribution, refers to a set of continuous probability distributions defined in the interval (0,1).
[0076] Gamma distribution is a continuous probability function in statistics, which contains two parameters α and β, where α is called the shape parameter and β is called the scale parameter.
[0077] Accordingly, the target classification is extracted from the initial classification according to the reference number, such as Figure 8 As shown, including:
[0078] S1035. Sort the initial categories corresponding to the reference numbers in descending order of the reference numbers.
[0079] S1037. Extract target categories based on the sorting results.
[0080] Specifically, the TOP N of the sorting results may be selected as target categories. For example, if there are 20 initial categories in the sorting results, and N=5, then the top 5 initial categories are selected as target categories. N may be adaptively selected based on actual conditions.
[0081] In the embodiment of the present invention, a reference number can be generated based on user behavior data for ranking reference, and the statistical result of the reference number generation is consistent with the user behavior data, so that the initial classification that the user is interested in is selected as the target classification with a greater probability, and the initial classification that the user rarely has the opportunity to contact or is not very interested in is selected as the target classification with a lower probability, but it is not completely impossible to be selected as the target classification. In this way, the user's other interests are detected on the basis of taking into account the user's existing interests, so as to avoid the content recommended to the user becoming narrower and narrower.
[0082] S105. Obtain the corresponding target information in each target category to obtain a target information set.
[0083] Specifically, the target information corresponding to each target category is obtained to obtain a target information set, such as Fig. 9 As shown, including:
[0084] S1051. Obtain the first category and the second category in the target category; the first category is the category hit by the recalled information; the second category is the other categories in the target category except the first category.
[0085] Specifically, in the embodiment of the present invention, recall information can be obtained from various angles based on at least one recall strategy, including but not limited to a combination of one or more of hotspot recall, regional recall, interest recall, click sequence recall, and matrix decomposition recall.
[0086] Hotspot recall: Recall the latest global hotspot information for personalized information recommendation.
[0087] Geographical recall: Recall relevant information based on the user's permanent residence, hometown, and real-time geographic location information for personalized information recommendation.
[0088] Interest recall: Recall relevant information based on the user's historical click behavior for personalized information recommendation. Specifically, the interest recall can use collaborative filtering recall of ICF and UCF. ICF refers to Item Collaborative Filtering, which analyzes and recommends from the perspective of information. UCF refers to User Collaborative Filtering, which refers to finding people with similar interests from the user's perspective and making recommendations.
[0089] Matrix decomposition recall: users and items are represented as two-dimensional matrices, and the recommendation task is regarded as a matrix completion task, that is, the elements in the matrix that have not been recorded are filled based on the existing data in the matrix to obtain the recommendation results.
[0090] Click sequence recall: Recall based on the user's click sequence.
[0091] According to the recall strategy, at least one piece of information can be obtained, and each piece of information has its corresponding initial classification. The classification hit by the information obtained according to the recall strategy is the primary classification, and the other target classifications are the secondary classifications.
[0092] Taking the target classification including 10 initial classifications as an example, the target classification includes <Life><International><Domestic><Suzhou><Entertainment><Sports><Finance><Society><Jokes><Automobile>, and the classification hit by the information obtained by the recalled strategy is <Entertainment><Sports><Finance><Society><Jokes><Automobile>, then the secondary classification is <Life><International><Domestic><Suzhou>.
[0093] S1053. Select a first amount of information for each first category to obtain a first target information set, and select a second amount of information for each second category to obtain a second target information set.
[0094] The first number and the second number may be the same or different. The information in the first category may be selected from the information recalled based on the recall strategy to select information with a high degree of recommendation or popularity. The information in the second category may be selected based on the popularity of other users' access to the information in the second category, or based on actual conditions (such as real-time nature, importance).
[0095] S1055. Obtain a target information set by combining the first target information set and the second target information set.
[0096] S107. Extract information for recommendation from the target information set.
[0097] Specifically, the inherent order of the information in the target information set may be disrupted based on a preset algorithm, and then the information recommended by the user may be selected therefrom, or information for recommendation may be randomly extracted from the target information set. The embodiment of the present invention does not limit the amount of information for recommendation. The purpose of disrupting the inherent order or randomly extracting is to make each piece of information in the target information set be recommended with the same or similar probability, thereby ensuring the coverage of the information recommended to the user.
[0098] In the embodiment of the present invention, the target classification for exploring the user's interests is obtained by analyzing the user's behavior data, thereby realizing an exploratory recall. Fig.10 , which shows a schematic diagram of an information recommendation method in an embodiment of the present invention, which supplements the logic of exploratory recall on the basis of the inherent recall strategy, and disrupts or disperses the results based on the exploratory recall, so that the exploratory recall results also have the opportunity to be recommended to users.
[0099] Please refer to Fig.11 , which shows a schematic diagram of the embodiment of the present invention in a user terminal in a news scenario. Users often browse content classified as <Entertainment>, <Life>, and <Automobile>, but hot information about <Real Estate> can also be recommended to users. Not only in news scenarios, the embodiment of the present invention can also be widely used in various scenarios that require information recommendation, such as video, microblog, music, games, shopping, etc.
[0100] An information recommendation method disclosed in an embodiment of the present invention adds logic for exploring user interests, and can optimize user interests for each recommendation, so that users have the opportunity to see information in categories that have not been exposed, and achieves overall optimization modeling of various recall results, thereby improving the accuracy and coverage of information recall. By continuously exploring user interests, the diversity and richness of user recommended information can be broadened, avoiding the problem of increasingly narrow coverage of recommended information.
[0101] The embodiment of the present invention also provides an information recommendation device, such as Fig.12 As shown, the device comprises:
[0102] A user behavior data or acquisition module 201 is used to acquire user behavior data, wherein the user behavior data is used to characterize the user's interest in various initial categories;
[0103] The user behavior data analysis module 203 is used to generate a reference number for each initial classification, and extract a target classification from the initial classification according to the reference number; the generation of the reference number conforms to a probability distribution, and the probability distribution is related to the user behavior data of the initial classification corresponding to it;
[0104] The target information set acquisition module 205 is used to acquire the target information corresponding to each target classification to obtain a target information set;
[0105] The recommendation information extraction module 207 is used to extract information for recommendation from the target information set.
[0106] The user behavior data smooth change module 209, the user behavior data smooth change module 209, Fig.13 As shown, including:
[0107] A target user behavior data pair determining unit 2091 is used to determine the target user behavior data pair pointed to by the user instruction;
[0108] A sum value calculation unit 2093, used to obtain the sum value of the selected value and the unselected value of the target user behavior data pair;
[0109] The target data to be changed determining unit 2095 is used to determine the target data to be changed in the target user behavior data pair according to the user instruction;
[0110] The first branch unit 2097 is configured to increment the data to be changed by one if the sum is not greater than a preset threshold;
[0111] The second branch unit 2099 is used for, if the sum value is greater than a preset threshold, increasing the data to be changed by one and then multiplying it by a smoothing coefficient related to the preset threshold.
[0112] Specifically, the information recommendation device and method embodiments described in the embodiments of the present invention are based on the same inventive concept.
[0113] The embodiment of the present invention further provides a computer storage medium, which can store multiple instructions, and the instructions are suitable for being loaded by a processor and executing various steps of an information recommendation method described in the embodiment of the present invention. The specific execution process includes:
[0114] An information recommendation method, the method comprising:
[0115] Acquire user behavior data, where the user behavior data is used to characterize the user's interest in various initial categories;
[0116] Generate a reference number for each initial classification, and extract a target classification from the initial classification according to the reference number; the generation of the reference number conforms to a probability distribution, and the probability distribution is related to the user behavior data of the initial classification corresponding to the reference number;
[0117] Obtaining the corresponding target information in each target classification to obtain a target information set;
[0118] Information for recommendation is extracted from the target information set.
[0119] Furthermore, before acquiring the user behavior data, the user behavior data is smoothly changed, and the user behavior data is smoothly changed, including:
[0120] In response to a user instruction, determining a target user behavior data pair pointed to by the user instruction;
[0121] Obtaining the sum of the selected value and the unselected value of the target user behavior data pair;
[0122] According to the user instruction, determining the target data to be changed in the target user behavior data pair;
[0123] If the sum is not greater than the preset threshold, the data to be changed is incremented by one;
[0124] If the sum is greater than a preset threshold, the data to be changed is incremented by one and then multiplied by a smoothing coefficient associated with the preset threshold.
[0125] Further, the generating of the reference number of each initial classification includes:
[0126] Obtaining a target user behavior data pair corresponding to each initial classification according to the user behavior data, wherein the target user behavior data pair includes a selected value and an unselected value;
[0127] For each initial classification, a reference number is generated based on its corresponding probability distribution, and the expected value of the probability distribution is related to the result of dividing the selected value by the sum of the selected value and the non-selected value in the target user behavior pair corresponding to the initial classification.
[0128] Further, extracting the target classification from the initial classification according to the reference number includes:
[0129] Sort each initial classification corresponding to the reference number in descending order of the reference number;
[0130] Extract the target classification based on the sorting results.
[0131] Further, the acquiring of the target information corresponding to each target classification to obtain a target information set includes:
[0132] Acquire a first category and a second category in the target category; the first category is the category hit by the recalled information; the second category is the other categories in the target category except the first category;
[0133] Selecting a first amount of information for each first category to obtain a first target information set, and selecting a second amount of information for each second category to obtain a second target information set;
[0134] The target information set is obtained by integrating the first target information set and the second target information set.
[0135] Further, extracting information for recommendation from the target information set includes:
[0136] Based on a preset algorithm, the inherent order of the information in the target information set is disturbed, and information recommended by the user is selected therefrom, or information for recommendation is randomly extracted from the target information set.
[0137] Further, Fig.14 A schematic diagram of the hardware structure of a device for implementing the method provided by an embodiment of the present invention is shown, and the device may participate in constituting or include the apparatus provided by an embodiment of the present invention. Fig.14As shown, the device 10 may include one or more (102a, 102b, ..., 102n are used to illustrate) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Fig.14 The structure shown is for illustration only and does not limit the structure of the electronic device. Fig.14 More or fewer components as shown, or with Fig.14 Different configurations are shown.
[0138] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the device 10 (or mobile device). As involved in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).
[0139] The memory 104 can be used to store software programs and modules of application software, such as program instructions / data storage devices corresponding to the method in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, to implement the above-mentioned information recommendation method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the device 10 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.
[0140] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the device 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (Radio Frequency, RF) module, which is used to communicate with the Internet wirelessly.
[0141] The display may be, for example, a touch screen liquid crystal display (LCD) that may enable a user to interact with a user interface of device 10 (or mobile device).
[0142] It should be noted that the sequence of the embodiments of the present invention described above is for description only and does not represent the advantages and disadvantages of the embodiments. The above describes specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0143] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and server embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.
[0144] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.
[0145] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. An information recommendation method, characterized in that: The method comprises: Acquire user behavior data, where the user behavior data is used to characterize the user's interest in various initial categories; According to the user behavior data, a target user behavior data pair corresponding to each initial classification is obtained, wherein the target user behavior data pair includes a selected value and a non-selected value; For each initial classification, a corresponding reference number is generated based on its corresponding probability distribution, wherein the expected value of the probability distribution is equal to or positively correlated to the result of dividing the selected value of the target user behavior pair corresponding to the initial classification by the sum of the selected value and the unselected value; Sort the initial categories corresponding to the reference numbers in descending order, and use the first preset number of initial categories as target categories; the reference numbers are used to make the probability of the initial category that the user is interested in being selected as the target category greater than the probability of other initial categories being selected as the target category, and the target category is used to complete the detection of other interests of the user on the basis of taking into account the existing interests of the user; Obtaining the corresponding target information in each target classification to obtain a target information set; Information for recommendation is extracted from the target information set.
2. The method according to claim 1, characterized in that: Before acquiring the user behavior data, the user behavior data is smoothly changed, and the user behavior data is smoothly changed, including: In response to a user instruction, determining a target user behavior data pair pointed to by the user instruction; Obtaining the sum of the selected value and the unselected value of the target user behavior data pair; According to the user instruction, determining the target data to be changed in the target user behavior data pair; If the sum is not greater than the preset threshold, the data to be changed is incremented by one; If the sum is greater than a preset threshold, the data to be changed is incremented by one and then multiplied by a smoothing coefficient associated with the preset threshold.
3. The method according to claim 1, characterized in that The step of obtaining the target information corresponding to each target classification to obtain a target information set includes: Acquire a first category and a second category in the target category; the first category is the category hit by the recalled information; the second category is the other categories in the target category except the first category; Selecting a first amount of information for each first category to obtain a first target information set, and selecting a second amount of information for each second category to obtain a second target information set; The target information set is obtained by integrating the first target information set and the second target information set.
4. The method according to claim 1, characterized in that: The extracting information for recommendation from the target information set includes: Based on a preset algorithm, the inherent order of the information in the target information set is disturbed, and information recommended by the user is selected therefrom, or information for recommendation is randomly extracted from the target information set.
5. An information recommendation device, characterized in that: The device comprises: A user behavior data acquisition module, used to acquire user behavior data, wherein the user behavior data is used to characterize the user's interest in various initial categories; A user behavior data analysis module is used to obtain a target user behavior data pair corresponding to each initial classification according to the user behavior data, wherein the target user behavior data pair includes a selected value and an unselected value; for each initial classification, a corresponding reference number is generated based on its corresponding probability distribution, wherein the expected value of the probability distribution is equal to or positively correlated to the result of dividing the selected value in the target user behavior pair corresponding to the initial classification by the sum of the selected value and the unselected value; the initial classifications corresponding to the reference numbers are sorted in descending order of the reference numbers, and the first preset number of initial classifications are used as target classifications; the reference number is used to make the probability of the initial classification of interest to the user being selected as the target classification greater than the probability of other initial classifications being selected as the target classification, and the target classification is used to complete the detection of other interests of the user on the basis of taking into account the existing interests of the user; A target information set acquisition module is used to acquire the target information corresponding to each target classification to obtain a target information set; The recommendation information extraction module is used to extract information for recommendation from the target information set.
6. The device according to claim 5, characterized in that It also includes a user behavior data smooth change module, the user behavior data smooth change module includes: a target user behavior data pair determining unit, configured to determine the target user behavior data pair pointed to by the user instruction; A sum value calculation unit, used to obtain the sum value of the selected value and the unselected value of the target user behavior data pair; a target data to be changed determining unit, configured to determine the target data to be changed in the target user behavior data pair according to the user instruction; A first branch unit, configured to increment the data to be changed by one if the sum value is not greater than a preset threshold; The second branch unit is used for, if the sum value is greater than a preset threshold, increasing the data to be changed by one and then multiplying it by a smoothing coefficient related to the preset threshold.
7. A device, characterized in that: The device includes a processor and a memory, wherein the memory stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement an information recommendation method according to any one of claims 1 to 4.
8. A computer storage medium, characterized in that: The storage medium stores at least one instruction, and the at least one instruction is loaded by the processor and executed by the information recommendation method according to any one of claims 1 to 4.
