IPTV program product packaging recommendation method

Through the intelligent IPTV program product packaging and recommendation method, using feature extraction and model recommendation, the automatic package and accurate push of IPTV platform content is realized, which solves the high cost and user experience problems of manual operation by operators, and improves operational efficiency and users Satisfaction.

CN118972647APending Publication Date: 2024-11-15PACO VIDEO TECH (HANGZHOU) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410937170.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-12
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

In the IPTV platform, operators who maintain large amounts of events and on-demand content need to manually add product packages, which leads to large workloads, time-consuming and labor-intensive work, and it is difficult to consider the interests and on-demand consumption habits of a large number of users, resulting in high operating costs and poor user experience. good.

Method used

The intelligent IPTV program product packaging recommendation method is adopted, and the feature extraction and input data of new content is obtained, and the target product package and user group is determined using the product packaging recommendation model to realize automatic package and accurate push of content.

Benefits of technology

Greatly reduce operational workload, improve user experience and platform operation effect, effectively reduce operation costs, and achieve accurate push and automatic package entry of new content.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118972647A_ABST
    Figure CN118972647A_ABST
Patent Text Reader

Abstract

The invention provides an IPTV (Internet Protocol Television) program product packaging recommendation method, which comprises the following steps: acquiring newly added IPTV program content which is not added into any existing IPTV program product package; performing feature extraction on the newly added IPTV program content, and determining input data corresponding to the newly added IPTV program content; inputting the input data corresponding to the newly added IPTV program content into an IPTV program product packaging recommendation model, and determining target product packages corresponding to the newly added IPTV program content and a user group corresponding to each target product package through the IPTV program product packaging recommendation model; and packaging the newly added IPTV program content into a target product package, and recommending the target product package to each user in the corresponding user group. The program content is packaged and accurately pushed in an intelligent mode, the operation cost is effectively reduced, and the user experience and the operation effect are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of IPTV technology, and in particular to a method for recommending IPTV program product packaging. Background Art

[0002] In the IPTV platform, all kinds of competition and on-demand content data must be added with corresponding product information according to the rules and marked with corresponding corner mark information. Since the platform focuses on competitions, the competitions are divided into live broadcast type and replay type. Of course, in addition to competitions, the platform also has on-demand content, and there is a certain correlation between on-demand videos and competitions. When the platform pushes content to users, it needs to take into account the user's interests, on-demand consumption habits, etc. Improving the accuracy of push and the on-demand rate of pushed content is beneficial to both users and the platform. Accurate push for users can allow users to quickly find content of interest and effectively reduce the probability of users missing content of interest. Accurate push is conducive to increasing the user's on-demand rate, thereby improving the platform's operating results. Therefore, achieving accurate content push is something that almost every platform needs to do.

[0003] The IPTV platform operated by our unit maintains more than 60 types of event content, hundreds of thousands of game information, and millions of on-demand content data. The traditional operation plan requires operators to add the corresponding product packages for each commentary and each on-demand data under each game. In order to quickly and accurately deliver the corresponding game to the corresponding product package, it is also necessary to consider different plans for different commentaries, and the related on-demand content must also be added to the corresponding product package. To achieve real-time updates, the workload is very large and the work content is complex. It is very time-consuming and labor-intensive for operations. It often takes eight operation groups to cover the preparation of so much content across the country. There are also some large-scale events with tight schedules. Since they are aimed at users across the country, wrong information is not allowed, which is very demanding for operations. Moreover, the production of many upstream contents is irregular, which requires a large number of operators to be on duty 24 hours a day. This operation plan has high labor costs and is difficult to take into account the interests and on-demand consumption habits of a large number of users. How to package new content and accurately push it to improve user experience and platform benefits under various conditions is a technical problem that needs to be solved in this field. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide an IPTV program product packaging recommendation method, so as to realize the packaging and accurate push of program content in an intelligent manner, effectively reduce operating costs, and improve user experience and operating results.

[0005] In order to achieve the above purpose, the embodiments of the present application are implemented in the following manner:

[0006] In a first aspect, an embodiment of the present application provides a method for recommending IPTV program product packaging, comprising: obtaining newly added IPTV program content, wherein the newly added IPTV program content is not added to any existing IPTV program product package; performing feature extraction on the newly added IPTV program content to determine input data corresponding to the newly added IPTV program content; inputting the input data corresponding to the newly added IPTV program content into an IPTV program product packaging recommendation model, and determining a target product package corresponding to the newly added IPTV program content and a user group corresponding to each target product package through the IPTV program product packaging recommendation model; packaging the newly added IPTV program content into a target product package, and recommending the target product package to each user in the corresponding user group.

[0007] In combination with the first aspect, in a first possible implementation method of the first aspect, the on-demand type of each IPTV program content is on-demand or non-on-demand, and has labels in multiple dimensions, and feature extraction is performed on the newly added IPTV program content to determine the input data corresponding to the newly added IPTV program content, including: feature extraction is performed on the newly added IPTV program content to extract on-demand type features and label features, wherein the on-demand type features are used to reveal whether the viewing of the IPTV program content requires on-demand viewing, and the label features include content label features, event label features, commentary label features, and broadcast type label features. The content label features are used to reveal the features of the IPTV program content in the content dimension, the event label features are used to reveal the features of the IPTV program content in the event dimension, the commentary label features are used to reveal the features of the IPTV program content in the commentary dimension, and the broadcast type label features are used to reveal whether the broadcast type of the IPTV program content is live or replay; the on-demand type features and label features are spliced ​​and merged to obtain the input data corresponding to the newly added IPTV program content.

[0008] In combination with the first possible implementation method of the first aspect, in a second possible implementation method of the first aspect, the IPTV program product packaging recommendation model includes an input unit, a product packaging recommendation unit and an output unit, wherein the input unit is used to input input data into the product packaging recommendation unit; the product packaging recommendation unit is used to determine a plurality of pending product packages based on the input data, and based on the pending product packages and the user data set, determine a plurality of target product packages and a user group corresponding to each target product package; and the output unit is used to output each target product package and the corresponding user group.

[0009] In combination with the second possible implementation manner of the first aspect, in a third possible implementation manner of the first aspect, the product packaging recommendation unit includes a product package module and a recommendation module, the product package module is used to obtain a product package set, and based on the product package set, determine a pending product package after pre-packaging the newly added IPTV program content corresponding to the input data; for each pending product package, the recommendation module is used to obtain a user data set, and calculate each user's on-demand index for the pending product package based on the user data set, and then determine the adaptation index of the pending product package based on each on-demand index, so as to determine a number of target product packages from all pending product packages based on the adaptation index of each pending product package, and determine that the user whose on-demand index corresponding to each target product package exceeds a set threshold is the target user corresponding to the target product package, so as to form a user group corresponding to the target product package.

[0010] In combination with the third possible implementation method of the first aspect, in a fourth possible implementation method of the first aspect, the product package module is specifically used to: obtain a product package set, wherein the product package set includes multiple product packages; filter the product package set to obtain a filtered product package subset; for each product package in the product package subset, pre-package using the newly added IPTV program content corresponding to the input data to obtain a pending product package after pre-package.

[0011] In combination with the third possible implementation method of the first aspect, in a fifth possible implementation method of the first aspect, the recommendation module calculates the on-demand index of each user for the pending product package in the following manner: extract features of the pending product package to determine input features corresponding to the pending product package; input the input features into an on-demand index prediction model to perform on-demand index prediction, and obtain the on-demand index of each user for the pending product package.

[0012] In combination with the third possible implementation method of the first aspect, in a sixth possible implementation method of the first aspect, the recommendation module calculates the on-demand index of each user for the pending product package in the following manner: extracting features of the pending product package to determine the product features of the pending product package; based on the user data set, determining the interest features and on-demand behavior feature function corresponding to each user data; for each user data: calculating the interest overlap between the product features and the interest features, and calculating the on-demand index based on the interest overlap and the on-demand behavior feature function.

[0013] In combination with the sixth possible implementation manner of the first aspect, in a seventh possible implementation manner of the first aspect, the feature value of each dimension in the product feature is used to reveal the weight of the feature in the product package, and the feature value of each dimension in the interest feature is used to reveal the weight of the feature in the user's interest, and the interest overlap between the product feature and the interest feature is calculated, including: based on the product feature and the interest feature, determining the overlap features of the two; based on the overlap features of the product feature and the interest feature, calculating the interest overlap in the following manner:

[0014]

[0015] Among them, C i Indicates the overlap between the interest feature of the i-th user data and the product feature. ij represents the jth interest feature value in the overlapped features between the interest feature of the i-th user data and the product feature, Pr ij It represents the jth product feature value among the overlapping features between the interest features and product features of the i-th user data, and n is the number of overlapping features between the interest features and product features of the i-th user data.

[0016] In combination with the seventh possible implementation manner of the first aspect, in an eighth possible implementation manner of the first aspect, calculating the on-demand index based on the interest overlap and the on-demand behavior characteristic function includes: calculating the on-demand index corresponding to the i-th user data using the following formula:

[0017]

[0018] Among them, V i represents the on-demand index corresponding to the i-th user data, revealing the probability that the user corresponding to the i-th user data will on-demand the pending product package, α i is the optimization parameter corresponding to the i-th user data, which is used to reveal the on-demand consumption habits of the user, α i ≥0.

[0019] In combination with the eighth possible implementation manner of the first aspect, in a ninth possible implementation manner of the first aspect, the recommendation module determines the adaptation index of the undetermined product package based on each on-demand index in the following manner:

[0020]

[0021] Wherein, A is the adaptation index of the pending product package, and N is the number of user data in the user data set.

[0022] Beneficial effects:

[0023] 1. When new program content is generated upstream, the newly added IPTV program content can be obtained (the newly added IPTV program content is not added to any existing IPTV program product package), and the features of the newly added IPTV program content are extracted to determine the input data corresponding to the newly added IPTV program content, and then the input data is input into the IPTV program product packaging recommendation model. The target product package corresponding to the newly added IPTV program content and the user group corresponding to each target product package are determined through the IPTV program product packaging recommendation model; the newly added IPTV program content is packaged into the target product package, and the target product package is recommended to each user in the corresponding user group. In this way, the IPTV program product packaging recommendation model is used for intermediate processing, and finally the target product package corresponding to the newly added IPTV program content and the user group corresponding to each target product package are determined, which can greatly reduce the workload of operations, realize automatic packaging and accurate push of new content, and improve the operation effect of the platform.

[0024] 2. Extract features of newly added IPTV program content, extract on-demand type features and label features (including content label features, event label features, commentary label features, broadcast type label features, etc.) to form input data, and then use the product package module of the product package recommendation unit in the IPTV program product package recommendation model to pre-enter the package (certain screening can be performed, for example, screening according to event labels) to obtain multiple pending product packages, and then use the recommendation module to predict each pending product package: calculate the on-demand index of each user for the pending product package based on the user data set, and then determine the adaptation index of the pending product package based on each on-demand index, so as to determine several target product packages from all pending product packages based on the adaptation index of each pending product package, so as to determine the users whose on-demand index corresponding to each target product package exceeds the set threshold as the target users corresponding to the target product package, and form a user group corresponding to the target product package. This approach can take into account the impact of adding new IPTV program content to a certain product package, user interests, and the probability of users ordering this product package. After comprehensive evaluation, the target product package and the corresponding target user group can be determined, thereby achieving automatic packaging and precise push of new IPTV program content, thereby improving platform operation results.

[0025] 3. When predicting the on-demand index of the user for the pending product package, a mature artificial intelligence model (such as decision tree, support vector machine, random forest, etc.) can be used. This part of the work can be completed after providing the corresponding data training model. The advantage of this method is that it can provide corresponding training data for training according to different types of program content, thereby improving the applicability of this solution. However, in order to further improve the accuracy of the prediction, considering the characteristics of this scenario, the user's interest characteristics, and on-demand consumption habits (usually associated with interest characteristics, but for different interests, there are different on-demand consumption motivations), this solution designs the corresponding on-demand index calculation formula (divided into two parts, one interest overlap calculation, and the other on-demand index calculation based on interest overlap) through data analysis to better fit the user's on-demand consumption habits for the program content of interest, thereby improving the accuracy of the on-demand index prediction. The target product package and the corresponding user group determined based on this can also be more accurate, realizing the intelligent packaging and precise push of new IPTV program content, effectively reducing operating costs, improving user experience, and improving the operating effect of the platform.

[0026] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0028] Figure 1 A flowchart of an IPTV program product packaging recommendation method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0030] See also Figure 1 , Figure 1 The flowchart of a method for recommending IPTV program product packaging provided by an embodiment of the present application is as follows: The method for recommending IPTV program product packaging is applied to a server and includes steps S10, S20, S30 and S40.

[0031] When new IPTV program content is generated upstream, first, the server may execute step S10.

[0032] Step S10: Acquire newly added IPTV program content, wherein the newly added IPTV program content is not added to any existing IPTV program product package.

[0033] In this embodiment, the server can obtain newly added IPTV program content, and the newly added IPTV program content is not added to any existing IPTV program product package. For ease of description, this embodiment is introduced by taking a single IPTV program content as an example. The on-demand type of each IPTV program content is on-demand or non-on-demand, and has multiple dimensions of labels, such as content label dimension, event label dimension, commentary label dimension, broadcast type label dimension, etc. The content label dimension mainly reveals the characteristics of IPTV program content in the content dimension, such as lore, crushing, cooperation, confrontation, reversal, etc. The event label dimension is used to reveal the characteristics of IPTV program content in the event dimension, such as competition items (such as football, basketball, tennis, etc.), CBA, provincial teams, leagues, cups, championships, etc.; the commentary label dimension is mainly used to reveal the characteristics of IPTV program content in the commentary dimension, such as commentators, commentary styles, etc.; the broadcast type label dimension is used to reveal the broadcast type of IPTV program content, such as live broadcast, replay, etc.

[0034] After acquiring the newly added IPTV program content, the server may execute step S20.

[0035] Step S20: extracting features of the newly added IPTV program content to determine input data corresponding to the newly added IPTV program content.

[0036] In this embodiment, the server may perform feature extraction on the newly added IPTV program content to determine input data corresponding to the newly added IPTV program content.

[0037] Exemplarily, the server can perform feature extraction on the newly added IPTV program content, and extract on-demand type features and label features, wherein the on-demand type features are used to reveal whether the viewing of the IPTV program content requires on-demand viewing, and the label features include content label features, event label features, commentary label features, and broadcast type label features. The content label features are used to reveal the features of the IPTV program content in the content dimension, the event label features are used to reveal the features of the IPTV program content in the event dimension, the commentary label features are used to reveal the features of the IPTV program content in the commentary dimension, and the broadcast type label features are used to reveal whether the broadcast type of the IPTV program content is live or replay.

[0038] After determining the on-demand type features and the label features, the server can splice and merge the on-demand type features and the label features to obtain input data corresponding to the newly added IPTV program content.

[0039] After obtaining the input data corresponding to the newly added IPTV program content, the server may execute step S30.

[0040] Step S30: Inputting the input data corresponding to the newly added IPTV program content into the IPTV program product packaging recommendation model, and determining the target product package corresponding to the newly added IPTV program content and the user group corresponding to each target product package through the IPTV program product packaging recommendation model.

[0041] In this embodiment, for ease of understanding, the IPTV program product packaging recommendation model is first introduced here. The IPTV program product packaging recommendation model includes an input unit, a product packaging recommendation unit and an output unit.

[0042] The input unit is used to input input data into the product packaging recommendation unit.

[0043] The product package recommendation unit is used to determine a plurality of pending product packages based on input data, and to determine a plurality of target product packages and a user group corresponding to each target product package based on the pending product packages and the user data set.

[0044] Exemplarily, the product package recommendation unit may include a product package module and a recommendation module.

[0045] The product package module is used to obtain a product package set, and based on the product package set, determine a pending product package after pre-packaging the newly added IPTV program content corresponding to the input data.

[0046] For each pending product package, the recommendation module is used to obtain a user data set, and calculate each user's on-demand index for the pending product package based on the user data set, and then determine the adaptation index of the pending product package based on each on-demand index, so as to determine several target product packages from all pending product packages based on the adaptation index of each pending product package, and determine that the users whose on-demand index corresponding to each target product package exceeds a set threshold are the target users corresponding to the target product package, so as to form a user group corresponding to the target product package.

[0047] The output unit is used to output each target product package and the corresponding user group.

[0048] Then, after inputting the input data corresponding to the newly added IPTV program content into the IPTV program product packaging recommendation model, the input unit of the IPTV program product packaging recommendation model can input the input data into the product packaging recommendation unit.

[0049] The product package module in the product packaging recommendation unit can obtain a product package collection (i.e., a collection of existing product packages, such as an event basic package, XX service package, etc., where XX refers to the event project. In addition, the naming of the product package is set according to needs and is not limited. Moreover, there are many types of product packages, and product packages are involved in non-event fields, such as film and television service packages, television service packages, etc.), and then determine the pending product package after pre-packaging the newly added IPTV program content corresponding to the input data based on the product package collection.

[0050] For example, matching and screening are performed based on tags such as content and events, and existing product packages that are more suitable for inclusion are determined from the product package set. The newly added IPTV program content corresponding to the input data is pre-packaged into this part of the selected existing product packages (pre-packaging here can be understood as adding the newly added IPTV program content to this product package, but it is not the final packaging, because the number of packages for the same IPTV program content needs to be limited to ensure the quality of the product package, and it also needs to comply with some policy requirements), and the corresponding pending product package is obtained.

[0051] After obtaining the pending product package, the recommendation module in the product package recommendation unit can predict the on-demand index of the pending product package.

[0052] First, the recommendation module can obtain the user data set, which can be called from the user database of the platform. The collection and processing of user data (such as forming user portraits, extracting interest features, on-demand consumption habit features, etc.) is not the focus of this solution. This solution can directly call and obtain it. In the obtained user data set, each user data contains the user's interest features and on-demand behavior feature functions. Different users may have very different interest features or very similar ones. Users with similar interests also have different on-demand consumption habits. Some users may on-demand content that generates interest, but some users may not. However, generally speaking, the richer the content of interest contained in the product package, the higher the probability of on-demand content. It is generally not a linear relationship. Therefore, after analyzing the interests and consumption behaviors of most users, they are characterized by functions, but different users need to set an optimization parameter to adjust to better match the relationship between their interests and on-demand behaviors.

[0053] Accordingly, for each pending product package, the recommendation module may calculate each user's on-demand index for the pending product package based on the user data set.

[0054] This application provides two solutions to predict each user's on-demand index for the pending product package:

[0055] The first solution is to use an artificial intelligence model to extract features from the pending product package, determine the input features corresponding to the pending product package, and then input the input features into the on-demand index prediction model to predict the on-demand index and obtain each user's on-demand index for the pending product package. This solution requires the construction of an artificial intelligence model, such as through decision trees, support vector machines, random forests, etc. The model and construction process of this type of solution are relatively mature. It mainly requires building a model, establishing a training data set, and training the model. After the training is completed, the on-demand index prediction model can be obtained to predict the on-demand index.

[0056] However, this solution is more data-intensive (it requires higher training data, and the larger the amount of data, the more accurate the prediction is generally). Its advantage is that it is more portable. For example, the program data related to the game can be replaced with film and television program data such as TV series and movies for training, and then it can be easily migrated to the corresponding scenario for application.

[0057] As for the second solution, it is to construct an on-demand behavior characteristic function by analyzing the relationship between user interests and on-demand behaviors. Since the relationship between the interests and on-demand behaviors of most users conforms to similar characteristics, the higher the interest and the more things that match the interests, the more likely it is to generate on-demand behaviors, but the forms are different (for example, some users are in better conditions and will order anything they are slightly interested in, some users order things that are more in line with their interests, and some users rarely or even have no on-demand behaviors). Therefore, it is necessary to configure an optimization parameter for each user to adjust the corresponding on-demand behavior characteristic function, which is conducive to accurate prediction.

[0058] Therefore, the recommendation module can extract features from the pending product package and determine the product features of the pending product package (the feature value of each dimension in the product feature is used to reveal the weight of the feature in the product package, because the pending product package contains multiple program contents, and the label features of some program contents may overlap, and this part of the overlapping features cannot be ignored, so it needs to be reflected in the feature value and expressed through the feature value. The more times a feature appears, the larger the feature value. A simple solution can even directly use the number of times a feature appears as the feature value of the feature); then based on the user data set, determine the interest features corresponding to each user data (user data has a weight distribution corresponding to the interest features, which is a relatively mature technology) and the on-demand behavior feature function.

[0059] Then, for each piece of user data:

[0060] The recommendation module can calculate the interest overlap between product features and interest features, and then calculate the on-demand index based on the interest overlap and the on-demand behavior feature function.

[0061] Exemplarily, the recommendation module can determine the overlapping features of the product features and the interest features based on the product features and the interest features. Here, the overlapping features in the product features and the interest features can be determined in a form similar to one-hot encoding (but different from the traditional one-hot encoding because the value is not 1).

[0062] Afterwards, the interest overlap can be calculated based on the overlap between product features and interest features in the following way:

[0063]

[0064] Among them, C i Indicates the overlap between the interest feature of the i-th user data and the product feature. ij represents the jth interest feature value in the overlapped features between the interest feature of the i-th user data and the product feature, Pr ij It represents the jth product feature value among the overlapping features between the interest features and product features of the i-th user data, and n is the number of overlapping features between the interest features and product features of the i-th user data.

[0065] Thus, the interest overlap between product features and interest features can be calculated. Based on this, the recommendation module can calculate the on-demand index based on the interest overlap and the on-demand behavior feature function. For example, the on-demand index corresponding to the i-th user data can be calculated using the following formula:

[0066]

[0067] Among them, V i represents the on-demand index corresponding to the i-th user data, revealing the probability that the user corresponding to the i-th user data will on-demand the pending product package, α i is the optimization parameter corresponding to the i-th user data, which is used to reveal the on-demand consumption habits of the user, α i ≥0.

[0068] Thus, the on-demand index corresponding to the i-th user data is calculated, that is, the on-demand index of the user of the i-th user data for the pending product package. In this way, the on-demand index of each user data in the user data set for the pending product package can be calculated.

[0069] Afterwards, the adaptation index of the pending product package may be further determined based on each on-demand index. For example, the following formula may be used to calculate each on-demand index to determine the adaptation index of the pending product package:

[0070]

[0071] Wherein, A is the adaptation index of the pending product package, and N is the number of user data in the user data set.

[0072] In this way, the adaptation index corresponding to each pending product package can be calculated. Considering the restrictions on the conditions for inclusion in the package, the part with the highest adaptation index can be selected as the target product package for the final inclusion of the program content in the package.

[0073] Of course, this process can be that the recommendation module obtains set rules. For example, the rule stipulates that only a set number (for example, 5) of pending product packages can be selected as target product packages for inclusion. In this case, the 5 pending product packages with the highest adaptation index can be selected as target product packages. For another example, the benefits of the platform can be further considered, and weights can be assigned to the pending product packages. By calculating the expected value formed by users (recorded as target users) whose on-demand index exceeds a set threshold (for example, 0.5) and the pending product packages, the top 5 pending product packages can be selected as target product packages according to the expected value.

[0074] After determining the target product package, the recommendation module may determine the users whose on-demand index corresponding to each target product package exceeds a set threshold (eg, 0.5) as target users corresponding to the target product package, thereby forming a user group corresponding to the target product package.

[0075] After the target product packages corresponding to the IPTV program contents and the user groups corresponding to each target product package are determined by using the IPTV program product package recommendation model, the server may execute step S40.

[0076] Step S40: Encapsulate the newly added IPTV program content into a target product package, and recommend the target product package to each user in the corresponding user group.

[0077] In this embodiment, the server can encapsulate the newly added IPTV program content into each target product package, and then recommend each target product package to each user in the corresponding user group, thereby realizing intelligent packaging and accurate recommendation of IPTV program content, effectively reducing operating costs, and improving user experience and platform operation effects.

[0078] In summary, the embodiment of the present application provides a method for recommending IPTV program product packaging. When new program content is generated upstream, the newly added IPTV program content can be obtained (the newly added IPTV program content is not added to any existing IPTV program product package), and the feature extraction of the newly added IPTV program content is performed to determine the input data corresponding to the newly added IPTV program content, and then the input data is input into the IPTV program product packaging recommendation model. The target product package corresponding to the newly added IPTV program content and the user group corresponding to each target product package are determined through the IPTV program product packaging recommendation model; the newly added IPTV program content is packaged into the target product package, and the target product package is recommended to each user in the corresponding user group. In this way, the IPTV program product packaging recommendation model is used to perform intermediate processing, and finally the target product package corresponding to the newly added IPTV program content and the user group corresponding to each target product package are determined, which can greatly reduce the workload of operations, realize automatic packaging and accurate push of new content, and improve the operation effect of the platform.

[0079] Feature extraction is performed on the newly added IPTV program content to extract on-demand type features and label features (including content label features, event label features, commentary label features, broadcast type label features, etc.) to form input data, and then the product package module of the product package recommendation unit in the IPTV program product package recommendation model is used to pre-enter the package (certain screening can be performed, for example, screening according to event labels) to obtain multiple pending product packages, and then the recommendation module is used to predict each pending product package: based on the user data set, the on-demand index of each user for the pending product package is calculated, and then the adaptation index of the pending product package is determined based on each on-demand index, so as to determine a number of target product packages from all the pending product packages based on the adaptation index of each pending product package, so as to determine the users whose on-demand index corresponding to each target product package exceeds the set threshold as the target users corresponding to the target product package, and form a user group corresponding to the target product package. This approach can take into account the impact of adding new IPTV program content to a certain product package, user interests, and the probability of users ordering this product package. After comprehensive evaluation, the target product package and the corresponding target user group can be determined, thereby achieving automatic packaging and precise push of new IPTV program content, thereby improving platform operation results.

[0080] When predicting the on-demand index of users for a pending product package, mature artificial intelligence models (such as decision trees, support vector machines, random forests, etc.) can be used. This part of the work can be completed after providing the corresponding data training model. The advantage of this method is that it can provide corresponding training data for training according to different types of program content, thereby improving the applicability of this solution. However, in order to further improve the accuracy of the prediction, considering the characteristics of this scenario, the user's interest characteristics, and on-demand consumption habits (usually associated with interest characteristics, but for different interests, there are different on-demand consumption motivations), this solution designs the corresponding on-demand index calculation formula (divided into two parts, one interest overlap calculation, and the other on-demand index calculation based on interest overlap) through data analysis to better fit the user's on-demand consumption habits for the program content of interest, thereby improving the accuracy of on-demand index prediction. The target product package and the corresponding user group determined based on this can also be more accurate, realizing the intelligent packaging and precise push of new IPTV program content, effectively reducing operating costs, improving user experience, and improving the operating effect of the platform.

[0081] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0082] The above description is only an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for recommending IPTV program product packaging, characterized in that: include: Acquire newly added IPTV program content, wherein the newly added IPTV program content is not added to any existing IPTV program product package; Extract features of newly added IPTV program content to determine input data corresponding to the newly added IPTV program content; Inputting the input data corresponding to the newly added IPTV program content into the IPTV program product packaging recommendation model, and determining the target product packages corresponding to the newly added IPTV program content and the user groups corresponding to each target product package through the IPTV program product packaging recommendation model; The newly added IPTV program content is packaged into a target product package, and the target product package is recommended to each user in the corresponding user group.

2. The IPTV program product packaging recommendation method according to claim 1, characterized in that: The on-demand type of each IPTV program content is on-demand or non-on-demand, and has multiple-dimensional labels. Feature extraction is performed on the newly added IPTV program content to determine the input data corresponding to the newly added IPTV program content, including: Feature extraction is performed on the newly added IPTV program content to extract on-demand type features and label features, wherein the on-demand type feature is used to reveal whether the viewing of the IPTV program content requires on-demand viewing, and the label feature includes content label features, event label features, commentary label features, and broadcast type label features. The content label feature is used to reveal the features of the IPTV program content in the content dimension, the event label feature is used to reveal the features of the IPTV program content in the event dimension, the commentary label feature is used to reveal the features of the IPTV program content in the commentary dimension, and the broadcast type label feature is used to reveal whether the broadcast type of the IPTV program content is live broadcast or playback; The on-demand type features and label features are spliced ​​and fused to obtain the input data corresponding to the newly added IPTV program content.

3. The IPTV program product packaging recommendation method according to claim 2, characterized in that: The IPTV program product packaging recommendation model includes an input unit, a product packaging recommendation unit and an output unit. The input unit is used to input the input data into the product packaging recommendation unit; The product package recommendation unit is used to determine a plurality of pending product packages based on the input data, and determine a plurality of target product packages and a user group corresponding to each target product package based on the pending product packages and the user data set; The output unit is used to output each target product package and the corresponding user group.

4. The IPTV program product packaging recommendation method according to claim 3, characterized in that: The product package recommendation unit includes a product package module and a recommendation module. The product package module is used to obtain a product package set, and based on the product package set, determine a pending product package after pre-packaging the newly added IPTV program content corresponding to the input data; For each pending product package, the recommendation module is used to obtain a user data set, and calculate each user's on-demand index for the pending product package based on the user data set, and then determine the adaptation index of the pending product package based on each on-demand index, so as to determine several target product packages from all the pending product packages based on the adaptation index of each pending product package, and determine that the users whose on-demand indexes corresponding to each target product package exceed a set threshold are the target users corresponding to the target product package, so as to form a user group corresponding to the target product package.

5. The IPTV program product packaging recommendation method according to claim 4, characterized in that: The product package module is specifically used for: Get a product package set, where the product package set contains multiple product packages; Filter the product package set to obtain a filtered product package subset; For each product package in the product package subset, pre-packaging is performed using the newly added IPTV program content corresponding to the input data to obtain a pending product package after pre-packaging.

6. The IPTV program product packaging recommendation method according to claim 4, characterized in that: The recommendation module calculates each user's on-demand index for the pending product package in the following manner: Perform feature extraction on the pending product package to determine the input features corresponding to the pending product package; The input features are input into the on-demand index prediction model to perform on-demand index prediction and obtain the on-demand index of each user for the pending product package.

7. The IPTV program product packaging recommendation method according to claim 4, characterized in that: The recommendation module calculates each user's on-demand index for the pending product package in the following manner: Extract features of the pending product package to determine product features of the pending product package; Based on the user data set, determine the interest characteristics and on-demand behavior characteristic functions corresponding to each user data; For each piece of user data: calculate the interest overlap between product features and interest features, and calculate the on-demand index based on the interest overlap and on-demand behavior feature function.

8. The IPTV program product packaging recommendation method according to claim 7, characterized in that: The feature value of each dimension in the product feature is used to reveal the weight of the feature in the product package. The feature value of each dimension in the interest feature is used to reveal the weight of the feature in the user's interest. The overlap between the product feature and the interest feature is calculated, including: Based on product features and interest features, determine the overlapping features of the two; Based on the overlap between product features and interest features, the interest overlap is calculated in the following way: Among them, C i Indicates the interest overlap between the interest feature of the i-th user data and the product feature. ij represents the jth interest feature value in the overlapped features between the interest feature of the i-th user data and the product feature, Pr ij It represents the jth product feature value among the overlapping features between the interest features and product features of the i-th user data, and n is the number of overlapping features between the interest features and product features of the i-th user data.

9. The IPTV program product packaging recommendation method according to claim 8, characterized in that: Based on the interest overlap and on-demand behavior characteristic function, the on-demand index is calculated, including: The on-demand index corresponding to the i-th user data is calculated using the following formula: Among them, V i represents the on-demand index corresponding to the i-th user data, revealing the probability that the user corresponding to the i-th user data will on-demand the pending product package, α i is the optimization parameter corresponding to the i-th user data, which is used to reveal the on-demand consumption habits of the user, α i ≥0.

10. The IPTV program product packaging recommendation method according to claim 9, characterized in that: The recommendation module determines the adaptation index of the pending product package based on each on-demand index in the following manner: Wherein, A is the adaptation index of the pending product package, and N is the number of user data in the user data set.