Program recommendation method, device and computer readable storage medium
By combining unsupervised and supervised recommendation methods, using the recommendation prediction model to supervise learning the prediction success rate of the target program, the problem of poor user experience in the existing technology is solved, and the accuracy of program recommendation and user satisfaction are improved.
Patent Information
- Application Number
- CN201911114300.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2019-11-14
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2039-11-14
AI Technical Summary
The existing program recommendation methods are difficult to accurately determine whether the user likes the recommended programs, resulting in poor user experience and may recommend programs that the user does not like or cannot watch.
Using a program recommendation method combining unsupervised and supervised recommendation methods, a recommendation prediction model is established by obtaining the recommended success rate and associated parameters of the recommended program, and the model is used to supervise the learning of the predicted success rate of the target program, when the predicted success rate reaches Recommend the target program when the threshold is preset, correct unsupervised recommendation results, and improve the accuracy of recommendations.
It increases the probability of recommended programs being watched, reduces the recommendations of programs that users dislike, and improves the user experience.
Smart Images

Figure CN112804566B_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present application relate to but are not limited to the field of information processing technology, and in particular, to a program recommendation method, device, and computer-readable storage medium. Background Art
[0002] With the rapid development of IPTV (Internet Protocol Television, interactive network television), the program formats and contents are becoming more and more abundant. How to recommend these abundant programs to IPTV users according to their preferences has become a hot research area. The current program recommendation method generally only clusters users with the same preferences or similar programs, and recommends related programs to groups with the same preferences or recommends similar programs to users. However, the existing recommendation method has the following problems: since it is unclear whether the user likes the recommended program, there is a problem of recommending programs that the user does not like or will not watch at all, which will cause a poor user experience. Summary of the invention
[0003] The following is a summary of the subject matter described in detail herein. This summary is not intended to limit the scope of the claims.
[0004] In a first aspect, the embodiments of the present application provide a program recommendation method, device, and computer-readable storage medium, which can increase the probability of the recommended programs being watched, thereby improving the user experience.
[0005] In a second aspect, an embodiment of the present application provides a program recommendation method, comprising:
[0006] Obtain the recommendation success rate and recommendation-related parameters of the recommended program;
[0007] Obtaining a recommendation prediction model according to the recommendation success rate and the recommendation associated parameters;
[0008] Acquire target association parameters of a target program, and obtain a prediction success rate of the target program according to the target association parameters and the recommended prediction model;
[0009] When the prediction success rate reaches a preset threshold, the target program is recommended.
[0010] In a third aspect, an embodiment of the present application further provides a device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the program recommendation method according to the second aspect described above is implemented.
[0011] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute the program recommendation method as described above.
[0012] The embodiment of the present application includes: obtaining a recommendation prediction model according to the obtained recommendation success rate and recommendation association parameters of the recommended program, the recommendation prediction model belongs to a supervised recommendation method, and the recommendation prediction model can be trained with the existing recommended programs to improve the recommendation prediction model; then, when it is necessary to recommend a program, the target association parameters of the target program are supervised and learned using the recommendation prediction model, so as to obtain the prediction success rate of the target program, and when the prediction success rate reaches a preset threshold, it indicates that the probability that the target program will be watched by the user is very high, so the target program is recommended. According to the scheme provided by the embodiment of the present application, through the coordinated processing of the recommendation prediction model and the existing recommended programs, the recommendation prediction model can be used for supervised learning of the existing recommended programs, and by analyzing the probability of the recommended programs being watched by the user, the final recommendation result can be corrected, thereby reducing the recommendation of programs that the user does not like, and achieving the purpose of increasing the probability of the recommended programs being watched, thereby improving the user experience.
[0013] Other features and advantages of the present application will be described in the following description, and partly become apparent from the description, or understood by practicing the present application. The purpose and other advantages of the present application can be realized and obtained by the structures specifically pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] The accompanying drawings are used to provide further understanding of the technical solution of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the technical solution of the present application and do not constitute a limitation on the technical solution of the present application.
[0015] Figure 1 It is a schematic diagram of a framework of a system structure platform for executing a program recommendation method provided by an embodiment of the present application;
[0016] Figure 2 is a flow chart of a program recommendation method provided by an embodiment of the present application;
[0017] Figure 3 is a flow chart of obtaining a recommendation success rate and recommendation-related parameters in a program recommendation method provided in another embodiment of the present application;
[0018] Figure 4 is a schematic diagram of a state machine provided by an embodiment of the present application;
[0019] Figure 5is a flow chart of obtaining a recommendation prediction model in a program recommendation method provided in another embodiment of the present application;
[0020] Figure 6 is a flowchart of recommending a target program in a program recommendation method provided by another embodiment of the present application;
[0021] Figure 7 It is a schematic diagram of a device provided by an embodiment of the present application. DETAILED DESCRIPTION
[0022] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is 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 present application and are not used to limit the present application.
[0023] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims or the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.
[0024] The present application provides a program recommendation method, device and computer-readable storage medium. First, an unsupervised recommendation method is used to recommend programs. According to the obtained recommendation success rate and recommendation-related parameters of the recommended programs, a recommendation prediction model is obtained. The recommendation prediction model belongs to a supervised recommendation method. The recommendation results of the unsupervised recommendation method can be supervised and learned to improve the recommendation prediction model and improve the accuracy of the recommendation. Then, when it is necessary to recommend a program, the target-related parameters of the target program are supervised and learned using the recommendation prediction model to obtain the prediction success rate of the target program. When the prediction success rate reaches a preset threshold, it indicates that the probability that the target program will be watched by the user is very high, so the target program is recommended. Therefore, through the coordinated processing of the unsupervised recommendation method and the supervised recommendation method, the recommendation prediction model belonging to the supervised recommendation method can be used to supervise the recommendation results of the unsupervised recommendation method to correct the recommendation results of the unsupervised recommendation method, thereby reducing the recommendation of programs that users do not like, and achieving the purpose of increasing the probability of the recommended program being watched, thereby improving the user experience.
[0025] The embodiments of the present application are further described below in conjunction with the accompanying drawings.
[0026] like Figure 1 As shown, Figure 1 It is a schematic diagram of the framework of a system structure platform for executing a program recommendation method provided by an embodiment of the present application.
[0027] like Figure 1 As shown, the system structure platform 100 includes a terminal 110 and a service device 120 that can communicate with each other. The terminal 110 includes a first memory 111, a first processor 112 and a first communication module 113. The service device 120 includes a second memory 121, a second processor 122 and a second communication module 123. The first communication module 113 is electrically connected to the first processor 112, and the first memory 111 and the first processor 112 can be connected via a bus or other methods. Figure 1 The second communication module 123 is electrically connected to the second processor 122, and the second memory 121 and the second processor 122 can be connected via a bus or other means. Figure 1 The example of connecting through bus is taken in the following.
[0028] In one embodiment, the terminal 110 may be a smart phone, tablet computer, personal computer, smart TV, network set-top box, etc., which is installed with an IPTV client and can provide users with programs that users like. When the terminal 110 is a network set-top box, it needs to cooperate with a TV set so as to provide users with programs that users like. The service device 120 may be a local server, a local area network server, or a cloud server, etc., or a virtual host that can build a server may be selected. It can obtain attribute data related to the program through the IPTV client installed in the terminal 110, and recommend programs that users like to the user based on these attribute data.
[0029] The first communication module 113 and the second communication module 123 can be connected wirelessly or by wire. For example, when the terminal 110 is a mobile smart device such as a smart phone or a tablet computer, the first communication module 113 in the terminal 110 and the second communication module 123 in the service device 120 can be connected wirelessly; when the terminal 110 is a device such as a personal computer, a smart TV or a network set-top box, the first communication module 113 in the terminal 110 and the second communication module 123 in the service device 120 can be connected wirelessly or by wire. When the first communication module 113 and the second communication module 123 are connected wirelessly, the first communication module 113 and the second communication module 123 can perform data transmission through a WIFI network or through a mobile network, wherein the mobile network can be a 3G network, an LTE network or a 5G network.
[0030] The service device 120 can obtain relevant record data of programs that the user has watched through the user log in the IPTV client in the terminal 110, and can recommend similar programs to the user based on the relevant record data using an unsupervised recommendation method. When the user watches or collects programs recommended by the service device 120, the user log of the IPTV client will record the relevant record data of these recommended programs. The service device 120 can establish a recommendation prediction model based on the relevant record data of these recommended programs, and can use the relevant record data of these recommended programs to train the recommendation prediction model to complete the supervised learning of the recommendation success rate of the recommended programs by the recommendation prediction model. When the prediction success rate for the target program output by the recommendation prediction model reaches a preset threshold, it can be considered that the recommendation prediction model can more accurately recommend the user's favorite programs to the user than the unsupervised recommendation method. At this time, the program can be recommended to the user based on the output result of the recommendation prediction model. Since the unsupervised recommendation method and the recommendation prediction model can cooperate with each other, the recommendation prediction model belonging to the supervised recommendation method can be used to supervise the recommendation results of the unsupervised recommendation method to correct the recommendation results of the unsupervised recommendation method. Therefore, it is possible to reduce the recommendation of programs that users do not like, thereby achieving the purpose of increasing the probability of the recommended programs being watched, thereby improving the user experience.
[0031] The first memory 111 and the second memory 121 are respectively used as a non-transitory computer-readable storage medium, which can be used to store non-transitory software programs and non-transitory computer executable programs. In addition, the first memory 111 and the second memory 121 may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the first memory 111 may optionally include a memory remotely arranged relative to the first processor 112, and these remote memories may be connected to the terminal 110 via a network. In other embodiments, the second memory 121 may optionally include a memory remotely arranged relative to the second processor 122, and these remote memories may be connected to the service device 120 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.
[0032] Those skilled in the art will understand that Figure 1 The device structure shown in the figure does not constitute a limitation on the system structure platform 100, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.
[0033] exist Figure 1In the system structure platform 100 shown, the first processor 112 in the terminal 110 can be used to call the program recommendation program stored in the first memory 111 to implement the program recommendation method; or the second processor 122 in the service device 120 can be used to call the program recommendation program stored in the second memory 121 to implement the program recommendation method. In some embodiments, in order to save storage resources of the terminal 110, the program recommendation program can be executed only in the service device 120 to implement the program recommendation method.
[0034] Based on the above-mentioned system structure platform 100, various embodiments of the program recommendation method of the present application are proposed.
[0035] like Figure 2 As shown, Figure 2 This is a flowchart of a program recommendation method provided by an embodiment of the present application. The program recommendation method includes but is not limited to the following steps:
[0036] Step S100, recommending programs using an unsupervised recommendation method.
[0037] In one embodiment, the unsupervised recommendation method can adopt clustering, collaborative filtering or random recommendation methods. When the user watches the program through the IPTV client in the terminal, the user log in the IPTV client will record the relevant record data of the program watched by the user, such as the program name, program label and column record data. At this time, the service device can obtain these relevant record data through the user log, and find out the programs watched by other users with the same favorite programs based on these relevant record data, or find out other programs corresponding to the programs with these relevant record data, so as to make initial recommendations to the user for these found programs to meet the user's needs to watch the recommended programs.
[0038] It can be understood by those skilled in the art that unsupervised recommendation methods such as clustering, collaborative filtering or random recommendation are all commonly used methods in the art for recommending programs to users. Therefore, the specific technical principles of unsupervised recommendation methods such as clustering, collaborative filtering or random recommendation are not described in detail here. It is worth noting that the unsupervised recommendation method used in this embodiment is not limited to the clustering, collaborative filtering or random recommendation described above, and other unsupervised recommendation methods that can achieve the same technical effect can be used.
[0039] It is worth noting that the unsupervised recommendation method used in this embodiment for initial program recommendation can provide important program sample parameters for subsequent operation steps, so step S100 is one of the important steps associated with the subsequent operation steps.
[0040] Step S200, obtaining the recommendation success rate and recommendation-related parameters of the recommended program.
[0041] In one embodiment, the recommendation-related parameters are attribute parameters that can affect the recommendation of the program, including but not limited to one or more of the program type, the number of times the program has been viewed, the program popularity and the program time decay, which can be obtained through the user log in the IPTV client, wherein the program time decay reflects the degree to which the popularity of the program gradually cools down over time. The smaller the program time decay, the higher the popularity and the more popular the program is. Conversely, the larger the program time decay, the lower the popularity and the less popular the program is.
[0042] In one embodiment, the recommendation success rate of the recommended program can be calculated based on the user's related operations on the recommended program. For example, when the user watches the recommended program or collects the recommended program, it means that the recommended program may be consistent with the user's viewing preferences. Therefore, the user's viewing operation or collection operation can improve the recommendation success rate of the recommended program.
[0043] In one embodiment, the sample range used to calculate the recommendation success rate of the recommended program can be adjusted according to actual conditions, and is not specifically limited in this embodiment. For example, the recommendation success rate can be calculated for each user's related operations on the recommended program; for another example, the recommendation success rate can be calculated for the related operations of the recommended program by users in a certain area; for another example, the recommendation success rate can be calculated for the related operations of the recommended program by users globally.
[0044] In one embodiment, after executing step S100, the user log and recommendation record of the IPTV client will save the user's relevant record data on the recommended program. Based on these relevant record data, the recommendation success rate and recommendation-related parameters of the recommended program can be obtained, which can provide important sample parameters for subsequent operation steps, thereby providing necessary basic conditions for subsequent operation steps.
[0045] Step S300, obtaining a recommendation prediction model according to the recommendation success rate and recommendation associated parameters.
[0046] In one embodiment, the recommendation prediction model can have different implementation modes. For example, the recommendation prediction model can be a supervised learning model such as a support vector machine, a neural network, a decision tree or a random forest. It is worth noting that the recommendation prediction model used in this embodiment is not limited to the support vector machine, neural network, decision tree or random forest described above. Other supervised learning models that can achieve the same technical effect can be used.
[0047] In one embodiment, the recommendation success rate can be used as the dependent variable of the recommendation prediction model, and the recommendation association parameter can be used as the independent variable of the recommendation prediction model, so as to establish and train the recommendation prediction model. It is worth noting that the recommendation prediction model used in this embodiment is a regression model, which can show the significant relationship between the recommendation success rate and the recommendation association parameter.
[0048] Step S400, obtaining target association parameters of the target program, and obtaining the prediction success rate of the target program according to the target association parameters and the recommended prediction model.
[0049] In one embodiment, the target association parameters are attribute parameters that can affect the recommendation of the program, including but not limited to one or more of the program type, the number of times the program is viewed, the program popularity and the program time decay, which can be obtained through the user log in the IPTV client, wherein the program time decay reflects the degree to which the popularity of the program gradually cools down over time. The smaller the program time decay, the higher the popularity and the more popular the program is. Conversely, the larger the program time decay, the lower the popularity and the less popular the program is.
[0050] In one embodiment, after the recommendation prediction model completes the supervised learning training according to the recommendation success rate and the recommendation association parameter, the recommendation prediction model can be used to recommend programs to users. When the recommendation prediction model is used to recommend programs to users, the target program to be recommended can be selected first by using an unsupervised recommendation method according to the relevant record data in the user log, and the target association parameter corresponding to the target program can be obtained from the relevant record data in the user log. Then, the target association parameter is input into the recommendation prediction model to calculate the prediction success rate of the target program. Therefore, the corresponding target program can be judged according to the prediction success rate to determine whether it is a program that matches the user's viewing preferences, thereby increasing the probability of the recommended program being watched, thereby improving the user's experience.
[0051] Step S500: When the prediction success rate reaches a preset threshold, the target program is recommended.
[0052] In one embodiment, after the recommendation prediction model calculates the prediction success rate of the target program according to the target association parameters of the target program, it is necessary to determine whether the prediction success rate reaches a preset threshold. If the prediction success rate reaches the preset threshold, it means that the corresponding target program may be more in line with the user's viewing preferences, and the user's probability of watching or collecting the target program will be higher, so the target program can be recommended to the user; if the prediction success rate is less than the preset threshold, it means that the corresponding target program may not meet the user's viewing preferences, and the user's probability of watching or collecting the target program is very low, so the target program may not be recommended to the user. It is worth noting that the preset threshold can be set according to actual usage needs. For example, the preset threshold can be set to 80% or 90%, etc. This embodiment does not limit the specific value of the preset threshold.
[0053] In one embodiment, according to the above steps S100, S200, S300, S400 and S500, the service device first uses an unsupervised recommendation method to recommend programs, and obtains a recommendation prediction model trained by the recommendation success rate and the recommendation association parameters according to the obtained recommendation success rate and recommendation association parameters of the recommended programs; then, when it is necessary to recommend a program, the target association parameters of the target program are supervised and learned using the recommendation prediction model, so as to obtain the prediction success rate of the target program. When the prediction success rate reaches a preset threshold, it indicates that the probability that the target program will be watched or collected by the user is very high, so the target program can be recommended. Therefore, through the coordinated processing of the unsupervised recommendation method and the supervised recommendation method, the recommendation prediction model belonging to the supervised recommendation method can be used to supervise the recommendation results of the unsupervised recommendation method to correct the recommendation results of the unsupervised recommendation method, thereby reducing the recommendation of programs that the user does not like, and achieving the purpose of increasing the probability of the recommended program being watched, thereby improving the user experience.
[0054] like Figure 3 As shown, in one embodiment, step S200 includes but is not limited to the following steps:
[0055] Step S210, obtaining the effective viewing times, recommendation times and recommendation-related parameters of the recommended program;
[0056] Step S220, obtaining the recommendation success rate of the recommended program according to the effective viewing times and the recommended times.
[0057] In one embodiment, the recommendation success rate of the recommended program can be calculated by the ratio of the effective viewing times and the recommended times. Therefore, the calculation formula for the recommendation success rate of the recommended program can be P=N1 / N2, wherein P is the recommendation success rate of the recommended program, N1 is the effective viewing times of the recommended program, and N2 is the recommended times of the recommended program.
[0058] In one embodiment, the number of times a recommended program is recommended can be obtained from the recommendation record of the IPTV client, and the effective number of views of the recommended program can be calculated from the user log of the IPTV client. It is worth noting that the effective number of views can be calculated in different ways. For example, the effective number of views can be counted using a state machine; for another example, the effective number of views can be counted based on the user's behavior curve for the recommended program. When the state machine is used to count the effective number of views, different states can be assigned to the recommended program. When the user performs different operations on the recommended program, the state of the recommended program can be changed. When the state of the recommended program becomes a viewing state or a favorite state, it means that the recommended program has been effectively watched once, so the effective number of views of the recommended program can be counted. When counting the effective number of views based on the user's behavior curve for the recommended program, a characteristic curve of the recommended program can be formed based on the program time, viewing time, user viewing time period and user collection behavior, and then a user viewing curve is formed based on the user's normal viewing behavior, and a global curve is formed based on the global user's viewing behavior for the recommended program. Then, the correlation between the characteristic curve, the user viewing curve and the global curve is calculated. If the three curves are similar, the effective number of views can be counted once.
[0059] In one embodiment, the recommendation success rate of a recommended program is calculated based on the effective viewing times and the recommended times of the recommended program. Therefore, the recommendation success rate can match the viewing preferences of the user. The higher the recommendation success rate, the higher the probability that the user will watch or collect the recommended program. Therefore, by using the recommendation success rate as the training target value of the recommendation prediction model in the subsequent steps, and using the recommendation association parameters of the recommended program as the sample parameters of the recommendation prediction model, the recommendation prediction model can be better supervised and trained, thereby improving the recommendation prediction model.
[0060] like Figure 4 As shown, in another embodiment, the effective viewing times are counted through a state machine, and the state machine includes an effective viewing state. When the state of the recommended program reaches the effective viewing state within a preset time, the effective viewing times are recorded once.
[0061] In one embodiment, referring to Figure 4The state machine can be set with states such as entering viewing through recommendation, viewing, pausing viewing, valid viewing and invalid viewing. It is worth noting that the state defined for the program in the state machine can be customized according to the actual situation. This embodiment does not limit the program state defined in the state machine. When the state machine is used to count the number of valid views, the state parameters such as program duration, viewing duration, user viewing time period and collection operation can be obtained from the user log of the IPTV client first, and then the state of the program is tracked and converted according to these state parameters. For example, the state of the program is converted according to different trigger conditions or events. Specifically, the trigger condition or event can be that the viewing duration reaches a preset time threshold or the user collects the program, etc. For example, if the state of the program reaches the valid viewing state within the preset time threshold, then it can be considered that the program has been effectively watched once, so the number of valid views can be recorded once.
[0062] In one embodiment, the preset duration and the preset time threshold can be set according to actual usage needs. For example, the preset duration can be set to 10 seconds or 1 minute, and the preset time threshold can be set to 5 minutes or 10 minutes. This embodiment does not limit the specific values of the preset duration and the preset time threshold.
[0063] The following is an example description:
[0064] Reference Figure 4 According to different operations of users on programs, the programs are set to have the following five states: invalid viewing state, valid viewing state, viewing state entered through recommendation, viewing state and paused viewing state. When the user opens the program recommendation preview page of the IPTV client, it means that the user has received the program recommendation information. At this time, the state machine starts to track the status of the recommended program, that is, the recommended program at this time enters the viewing state entered through recommendation. When the user triggers such as Figure 4When the relevant conditions shown in , the status of the recommended program will be changed accordingly. For example, when the recommended program is in the viewing state, the background server of the IPTV client will accumulate the viewing time of the recommended program. If the viewing time is greater than or equal to the preset time threshold, the recommended program is triggered to enter the effective viewing state. For another example, when the user opens the program recommendation preview page, the status of the recommended program is in the viewing state through recommendation. If the user exits the program recommendation preview page, the status of the recommended program will be changed to the paused viewing state. At this time, if the user collects the recommended program, the recommended program will enter the effective viewing state; if the user does not collect the recommended program and does not continue to watch the recommended program within the preset time, the recommended program will enter the invalid viewing state. When the recommended program enters the effective viewing state, the state machine will stop tracking the status of the recommended program, and the effective viewing times of the recommended program will be effectively counted. Therefore, by accumulating the number of times the recommended program enters the effective viewing state, the effective viewing times of the recommended program can be obtained.
[0065] In one embodiment, the state of the recommended program is tracked through a state machine. As long as the recommended program enters the effective viewing state, the effective viewing times can be recorded once. Therefore, the state machine can be used to more conveniently record and count the effective viewing times, thereby improving the efficiency of subsequent accurate recommendation of programs.
[0066] like Figure 5 As shown, in another embodiment, step S300 includes but is not limited to the following steps:
[0067] Step S310, establishing a mapping relationship between recommendation success rate and recommendation associated parameters;
[0068] Step S320, taking the recommendation success rate as the training target value and the recommendation associated parameters as the training samples, training the mapping relationship to obtain a recommendation prediction model.
[0069] In one embodiment, a mapping relationship between the recommendation success rate and the recommendation-related parameters is established, which can be expressed by the following formula: Y=F(X), wherein Y is the recommendation success rate, which is a training target value; X is the recommendation-related parameter, which is a sample parameter used to train the recommendation prediction model; F() is the mapping relationship between the recommendation success rate Y and the recommendation-related parameter X, which belongs to the mapping relationship of nonlinear regression. It is worth noting that the formula Y=F(X) constitutes the recommendation prediction model. By inputting the recommendation success rate Y and the recommendation-related parameter X into the recommendation prediction model, the weight value of the recommendation prediction model can be trained so that the weight value of the recommendation prediction model can more accurately correspond to the mapping relationship between the recommendation success rate Y and the recommendation-related parameter X. The recommendation success rate Y and the recommendation-related parameter X can be matrixed first, and then input into the recommendation prediction model. In a specific example, the matrix composed of the recommendation success rate Y and the recommendation-related parameter X can be shown as follows:
[0070]
[0071] Among them, the first row of data in the matrix are the number of times the first recommended program has been viewed X11, the program time attenuation X12 of the first recommended program, the program popularity X13 of the first recommended program and the recommendation success rate Y1 of the first recommended program; the data types of the second row of data and the third row of data in the matrix are the same as the data types of the first row of data in the matrix, and the only difference between the three is that the corresponding recommended programs are different.
[0072] like Figure 6 As shown, in another embodiment, step S500 includes but is not limited to the following steps:
[0073] Step S510, when the prediction success rate reaches a preset threshold, a recommendation weight value is obtained based on the prediction success rate and the recommendation success rate;
[0074] Step S520: When the recommendation weight value reaches a preset weight threshold, the target program is recommended.
[0075] In one embodiment, the target program to be recommended can be comprehensively determined based on the recommendation results of the unsupervised recommendation method and the recommendation results of the recommendation prediction model. In order to balance the recommendation results of the two recommendation methods to ensure that the target program to be recommended meets the user's viewing preferences, a recommendation weight value can be introduced. If the recommendation weight value reaches the preset weight threshold, it means that the recommendation result of the recommendation prediction model is more in line with the user's viewing preferences. Therefore, the recommendation prediction model is used to recommend the program. If the recommendation weight value is less than the preset weight threshold, it means that the recommendation result of the unsupervised recommendation method is more in line with the user's viewing preferences. Therefore, the unsupervised recommendation method is used to recommend the program.
[0076] In one embodiment, since the recommendation prediction model belongs to a regression model, the recommendation weight value can be obtained based on the prediction success rate and the recommendation success rate using a regression model evaluation method. Among them, the regression model evaluation method can have different implementation methods. For example, the regression model evaluation method can use any one of SSE (The Sum of Squares due to Error), MSE (Mean Squared Error), RMSE (Root Mean Squared Error) and MAPE (Mean Absolute Percentage Error) to calculate the recommendation weight value. Taking the regression model evaluation method using SSE to calculate the recommendation weight value as an example, first calculate SSE. The calculation formula of SSE is as follows: SSE = ∑ (P1-P2) 2 , where P1 is the recommendation success rate, P2 is the prediction success rate, and the smaller the SSE value is, the closer the prediction success rate is to the actual recommendation success rate; then, the recommendation weight value is calculated according to the formula D=1 / SSE, where D is the recommendation weight value. When the recommendation weight value reaches the preset weight threshold, it means that the target program is more in line with the user's viewing preferences, so the target program is recommended to the user. It is worth noting that the preset weight threshold can be set according to actual usage needs, and this embodiment does not limit the specific value of the preset weight threshold.
[0077] In addition, in another embodiment, the number of target programs is two or more, and when the recommendation weight value reaches a preset weight threshold, the target programs are sorted based on the prediction success rate, recommendation success rate and recommendation weight value, and the target programs are recommended in order.
[0078] In one embodiment, when the number of target programs recommended to the user is two or more, there may be some target programs that do not meet the user's viewing preferences. Therefore, it is necessary to calculate the corresponding recommendation weight value for each target program, and then compare the recommendation weight value corresponding to each target program with the preset weight threshold, so that the target programs that meet the user's viewing preferences can be screened out. After the target programs that meet the user's viewing preferences are screened out, these target programs can be sorted according to the degree of closeness to the user's viewing preferences, and then these target programs can be recommended to the user in order. Therefore, when the user opens the program recommendation preview page of the IPTV client, it can be seen that the first recommended program is the program that best meets the user's viewing preferences, thereby satisfying the user's viewing needs and improving the user's user experience.
[0079] In one embodiment, the target programs are sorted by combining the prediction success rate, the recommendation success rate and the recommendation weight value, and the recommendation priority can be obtained according to the following formula: M = P 预 *P 推 *P 权 , where M is the recommended priority. The larger the value of M, the higher the priority. 预 is the prediction success rate; P 推 is the recommendation success rate; P 权 Next, the corresponding target programs are sorted according to the value of the recommendation priority M.
[0080] In addition, in another embodiment, the program recommendation method further includes the following steps:
[0081] Step S600: When the recommendation weight value is less than a preset weight threshold, an unsupervised recommendation method is used to recommend a program.
[0082] In one embodiment, when the recommendation weight value is less than a preset weight threshold, it means that the program recommended by the recommendation prediction model is more in line with the user's viewing preferences than the program recommended by the unsupervised recommendation method. Therefore, the unsupervised recommendation method is used for program recommendation.
[0083] In one embodiment, in the early stage of executing the program recommendation method, since the number of samples of recommendation success rate and recommendation-related parameters is relatively small, the recommendation prediction model cannot fully play its role. Therefore, an unsupervised recommendation method will be used to recommend programs first. When the number of samples of recommendation success rate and recommendation-related parameters is sufficient to enable the recommendation prediction model to complete supervised learning, the recommendation effects of the unsupervised recommendation method and the recommendation prediction model can be combined, so that the recommendation prediction model can correct the recommendation results of the unsupervised recommendation method, thereby recommending to users more accurate programs that meet the user's viewing preferences, thereby achieving the purpose of increasing the probability of the recommended programs being watched, thereby improving the user's experience.
[0084] like Figure 7 As shown, an embodiment of the present application provides a device, and the device 200 can be a local server, a local area network server or a cloud server, etc., and a virtual host that can build a server can also be selected.
[0085] Specifically, the device 200 includes: a memory 201 , a processor 202 , and a computer program stored in the memory 201 and executable on the processor 202 .
[0086] The processor 202 and the memory 201 may be connected via a bus or other means. Figure 7 The example of connecting through bus is taken in the following.
[0087] It should be noted that the device 200 in this embodiment is Figure 1 The system structure platform 100 in the embodiment shown in the figure is based on the same inventive concept. The device 200 in this embodiment can be constituted Figure 1 The system structure platform 100 in the illustrated embodiment is a part of the system structure platform 100, so both have the same implementation principles and beneficial effects, which will not be described in detail here.
[0088] The non-transient software program and instructions required to implement the program recommendation method in the above embodiment are stored in the memory 201. When executed by the processor 202, the program recommendation method in the above embodiment is executed, for example, the program recommendation method described above is executed. Figure 2 The method steps S100 to S500, Figure 3 Steps S210 to S220 of the method, Figure 5 Steps S310 to S320 of the method, Figure 6 Method steps S510 to S520 in.
[0089] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0090] In addition, an embodiment of the present application further provides a computer-readable storage medium, the computer-readable storage medium storing computer-executable instructions, the computer-executable instructions being executed by a processor or controller, for example, Figure 7 The processor 202 in the embodiment may be executed to enable the processor 202 to execute the program recommendation method in the above embodiment, for example, to execute the above described Figure 2 The method steps S100 to S500, Figure 3 Steps S210 to S220 of the method, Figure 5 Steps S310 to S320 of the method, Figure 6 Method steps S510 to S520 in.
[0091] It will be appreciated by those skilled in the art that all or some of the steps and systems in the disclosed method above may be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that may be used to store desired information and may be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.
[0092] The above is a specific description of the preferred implementation of the present application, but the present application is not limited to the above-mentioned implementation mode. Technical personnel familiar with the field can also make various equivalent deformations or substitutions without violating the spirit of the present application. These equivalent deformations or substitutions are all included in the scope defined by the claims of the present application.
Claims
1. A program recommendation method, comprising: Obtain the recommendation success rate and recommendation-related parameters of the recommended program; Establishing a mapping relationship between the recommendation success rate and the recommendation associated parameters; Taking the recommendation success rate as a training target value and the recommendation associated parameters as training samples, training the mapping relationship to obtain a recommendation prediction model; Acquire target association parameters of a target program, and obtain a prediction success rate of the target program according to the target association parameters and the recommended prediction model; When the prediction success rate reaches a preset threshold, a recommendation weight value is obtained based on the prediction success rate and the recommendation success rate of the target program, wherein the recommendation weight value is obtained based on the prediction success rate and the recommendation success rate of the target program, including: firstly calculating the sum variance SSE, the calculation formula of SSE is SSE=∑(P1-P2) 2 , where P1 is the recommendation success rate of the target program, and P2 is the prediction success rate of the target program; then the recommendation weight value is calculated according to the formula D=1 / SSE, where D is the recommendation weight value; In the case that the number of the target program is one, when the recommendation weight value reaches a preset weight threshold, the target program is recommended; In the case that the number of the target programs is two or more, when the recommendation weight value reaches a preset weight threshold, a recommendation priority is obtained according to the product of the prediction success rate, the recommendation success rate and the recommendation weight value, the target programs are sorted according to the recommendation priority, and the target programs are recommended in order; When the recommendation weight value is less than the preset weight threshold, an unsupervised recommendation method is used to recommend programs.
2. The program recommendation method according to claim 1, characterized in that: The obtaining of the recommendation success rate and recommendation-related parameters of the recommended program includes: Obtain the effective viewing times, recommended times and recommendation-related parameters of the recommended program; The recommendation success rate of the recommended program is obtained according to the effective viewing times and the recommended times.
3. The program recommendation method according to claim 2, characterized in that: The effective viewing times are counted by a state machine, wherein the state machine includes an effective viewing state. When the state of the recommended program reaches the effective viewing state within a preset time period, the effective viewing times are recorded once.
4. The program recommendation method according to claim 1, characterized in that: Before executing the step of obtaining the recommendation success rate and recommendation-related parameters of the recommended program, the following steps are performed: Program recommendation via unsupervised recommendation methods.
5. The program recommendation method according to any one of claims 1 to 4, characterized in that: The recommendation-related parameter and the target-related parameter respectively include at least one of the following: Type of program; Number of times the show has been viewed; Program popularity; Program time decay.
6. A device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the program recommendation method as described in any one of claims 1 to 5 when executing the computer program.
7. A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute the program recommendation method as described in any one of claims 1 to 5.
