Interactive network television service method and system

Abstract

The invention discloses an interactive network television service method and system, wherein the main method comprises: step one: constructing a recommendation matching effective item set by using the behavior data of IPTV users; step two: mining the implicit feedback of users in IPTV based on the constructed recommendation matching effective item set; step three: utilizing the viewing time of users and the total video time in the implicit connection to calculate the importance ranking of implicit factors that have an impact on the recommendation of video programs in the IPTV system; step four: calculating the selection rate of the user for the video program type of interest based on the mined implicit feedback data of users; step five: based on the existing historical viewing behavior data of users, expanding the recommendation of video programs that the user has never watched but meets the implicit needs of the user according to the behavior records of users with similar interests. The invention has the characteristics of improving the accuracy of system recommendations containing individual preferences and improving the payment conversion rate brought about by implicit needs.

Description

Technical Field

[0001] The present invention relates to the technical field of IPTV services, and in particular to an interactive network television service method and system. Background Art

[0002] IPTV, or Internet TV, is an Internet-based technology and a new media form of personalized, interactive services. It uses ADSL, Ethernet, or cable TV networks to access broadband networks, transmits TV signals through the Internet Protocol (IP protocol), and uses home TVs or computers as the main receiving terminals to provide a variety of digital media services including TV programs. Through IPTV applications, a variety of business forms including video on demand, live video, and Internet browsing can be realized. The existing IPTV service system analyzes and recommends interests based on the preference data generated by users in the daily use of IPTV, and then determines the service recommendations to be made in the next stage based on these recommendation effects. However, the interactive behaviors in the existing IPTV service system are less than those in video applications, resulting in a large number of implicit feedbacks, which makes it impossible for the system to distinguish the video content that users do not like or do not notice that they have not interacted with. Conventional recommendations will directly affect the recommendation accuracy and payment conversion rate of value-added services in the IPTV service system. Therefore, it is necessary to design an interactive network TV service method and system to improve the accuracy of service recommendations containing individual preferences and the payment conversion rate brought about by implicit demands. Summary of the invention

[0003] The object of the present invention is to provide an interactive network television service method and system to solve the problems raised in the above background technology.

[0004] In order to solve the above technical problems, the present invention provides the following technical solutions: an interactive network television service method, comprising the following steps:

[0005] Step 1: Use IPTV user behavior data to construct a valid item set for recommendation matching;

[0006] Step 2: Mining users’ implicit feedback in IPTV based on the constructed recommendation matching valid item set;

[0007] Step 3: Using the user's viewing time and the total video time in the implicit connection, calculate the importance ranking of the implicit factors that affect the recommendation of video programs in the IPTV system;

[0008] Step 4: Calculate the selection rate of the user's interested video program type based on the mined user implicit feedback data;

[0009] Step 5: Based on the user's existing historical viewing behavior data, according to the behavior records of users with similar interests, expand the recommendation of video programs that the user has never watched but meet the user's implicit needs.

[0010] According to the above technical solution, the method process steps for constructing a valid item set for recommendation matching include:

[0011] Step 11: Obtain the viewing log record data of the IPTV user, extract the event data corresponding to the user behavior state including the user ID, time, and channel ID field information, and establish the main feature set of the event data;

[0012] Step 12: Take the IPTV channel ID as the user's behavior item set, form the user's item sequence according to the order in which the behavior is generated, that is, the time sequence in which the user watches the channel, and construct a combination of the target event and the user-event interaction sequence. The target event is represented by its main features as the valid item set, and the user-event interaction is represented by a set of event features containing time codes;

[0013] Step 13: Expand the single mode to multi-mode from the user's item sequence, extract the scalable multi-mode item set to construct an effective feature set for recommendation matching, and take the order information of the user's viewing channels into consideration in feature extraction, considering not only the user features formed by a single channel, but also the connection between channels;

[0014] Step 14: The constructed recommended matching valid item set is transmitted to the IPTV data mining and analysis terminal, and the mining and analysis terminal performs analysis and mining based on the received three types of data including user data, program data, and user viewing log.

[0015] According to the above technical solution, the method and steps for mining implicit feedback of users in IPTV include:

[0016] Step 21: Establish a user interest mining model based on the received user data, program data, and user viewing log data, quantify the user viewing behavior, reflect the user interest through the user viewing behavior, and mine the viewing behavior that can stimulate the user;

[0017] Step 22: Synchronously extract the viewing time corresponding to the user's viewing record event, use the viewing time analysis to determine the user's interest in the current video, propose a user classification algorithm based on the user's viewing time to classify the user, and analyze the user's implicit viewing habits.

[0018] Step 23: Linking user interests and program types through IPTV program type labels, and building a label-based user viewing time sequence matrix based on the proportion of user viewing time in a specified time period, and extracting user implicit interest features using a clustering method based on a Gaussian mixture model;

[0019] Step 24: Utilizing the temporal characteristics of the viewing records of users in the IPTV system, the multi-modal item set is used as the identification input of the user interest mining model. Since IPTV is user-oriented, the viewing behavior of a user ID may reflect the interests of different individuals. The user's implicit interests are extracted using a recurrent neural network. An attention mechanism is added to the recurrent neural network model to distinguish the impact of different historical interest states on the current user's viewing decision.

[0020] Step 25: Label other users with similar target interests to the current user with corresponding tags and classify them into a total user implicit feedback set;

[0021] Step 26: Describe the implicit feedback characteristics of users in IPTV through annotated tags, use coarse-grained tags to represent the types of videos that users are interested in hidden in the IPTV system, and use fine-grained tags to represent the implicit connection between the types of videos that users are interested in and the system recommendations based on the tags.

[0022] According to the above technical solution, the method and steps for mining implicit feedback of users in IPTV include:

[0023] Step 21: Establish a user interest mining model based on the received user data, program data, and user viewing log data, quantify the user viewing behavior, reflect the user interest through the user viewing behavior, and mine the viewing behavior that can stimulate the user;

[0024] Step 22: Synchronously extract the viewing time corresponding to the user's viewing record event, use the viewing time analysis to determine the user's interest in the current video, propose a user classification algorithm based on the user's viewing time to classify the user, and analyze the user's implicit viewing habits.

[0025] Step 23: Linking user interests and program types through IPTV program type labels, and building a label-based user viewing time sequence matrix based on the proportion of user viewing time in a specified time period, and extracting user implicit interest features using a clustering method based on a Gaussian mixture model;

[0026] Step 24: Utilizing the temporal characteristics of the viewing records of users in the IPTV system, the multi-modal item set is used as the identification input of the user interest mining model. Since IPTV is user-oriented, the viewing behavior of a user ID may reflect the interests of different individuals. The user's implicit interests are extracted using a recurrent neural network. An attention mechanism is added to the recurrent neural network model to distinguish the impact of different historical interest states on the current user's viewing decision.

[0027] Step 25: Label other users with similar target interests to the current user with corresponding tags and classify them into a total user implicit feedback set;

[0028] Step 26: Describe the implicit feedback characteristics of users in IPTV through annotated tags, use coarse-grained tags to represent the types of videos that users are interested in hidden in the IPTV system, and use fine-grained tags to represent the implicit connection between the types of videos that users are interested in and the system recommendations based on the tags.

[0029] According to the above technical solution, the step of calculating the selection rate of the user for the video program type of interest includes:

[0030] The time period is set according to the mined implicit feedback data of users, and the unit of the time period is set to hours, each hour represents a time period. In the IPTV family group user recommendation scenario, the time period with the most appearances in all the user's viewing records is defined as the main time period T0, and k is the time period T0 obtained by calculating t 比值 Ratio to average The logarithm of is used to reflect whether the user's viewing interest exceeds the average value. For family users, the change in interest is mainly caused by the switching of members. The selection rate of the user in the specified time period is established based on the similarity of the viewing time periods of the members. That is, the user's selection rate is Q. By measuring the probability of switching behavior of family group users based on viewing time, the impact of member switching on users' interest preferences for current video programs is captured.

[0031] According to the above technical solution, the step of extending the recommendation includes:

[0032] Automatic clustering based on user behavior statistics. When different types of video programs are selected and clicked by multiple users at the same time, user interest characteristics and user classification based on viewing time statistics are used as the main influencing features of user selection rate. Video programs are divided into implicitly associated categories, namely light viewing, normal viewing, high viewing, and heavy viewing. The granularity of classification is controlled according to the implicitly associated categories. The more categories there are, the finer the granularity will be. The weight of each video program belonging to each category is calculated. When users who like a certain category will choose one of the video programs, then the weight of the video program in this category is higher. The video programs in the set that the target user has not subscribed to are sorted in descending order according to the subscription frequency and selection rate. According to this weight value, video media assets that meet the user's interests but have never been selected by the user are recommended to the user, forming a surprising content recommendation list of different regions based on the user's implicit feedback.

[0033] According to the above technical solution, the interactive network television service system includes:

[0034] The effective item set construction module is used to construct the effective item set for recommendation matching by using the behavior data of IPTV users;

[0035] Implicit feedback mining and analysis module, used to mine users' implicit feedback in IPTV based on the constructed recommendation matching valid item set;

[0036] The implicit feedback recommendation module is used to expand the recommendation of video programs that the user has never watched but meets the user's implicit needs.

[0037] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: the present invention constructs a valid item set for recommendation matching by utilizing the behavior data of IPTV users, and then mines the implicit feedback of users in IPTV based on the constructed valid item set for recommendation matching, and at the same time utilizes the viewing time of users and the total video time in implicit connections to perform importance ranking calculation of implicit factors that affect the recommendation of video programs in the IPTV system; then calculates the selection rate of the user for the type of video program that the user is interested in, so that based on the user's existing historical viewing behavior data, according to the behavior records of users with similar interests, the user is recommended to expand the video programs that the user has never watched but meet the user's implicit needs, and the user's implicit needs are mined according to the disassembly analysis, and based on this, the recommendation is expanded, which greatly improves the recommendation effect while also improving the accuracy of the recommendation based on individual preferences in the system, and further improves the characteristics of the payment conversion rate brought about by implicit needs in the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0039] Figure 1 A flowchart of an interactive network television service method provided by Embodiment 1 of the present invention;

[0040] Figure 2 This is a schematic diagram of the module composition of an interactive network television service system provided in the second embodiment of the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] Embodiment 1: Figure 1 This is a flowchart of an interactive network television service method provided by Embodiment 1 of the present invention. This embodiment can be applied to the recommendation service scenario of IPTV. This method can be executed by an interactive network television service system provided by this embodiment. Figure 1 As shown, the method specifically comprises the following steps:

[0043] Step 1: Use IPTV users’ behavior data to construct a valid item set for recommendation matching;

[0044] In an embodiment of the present invention, the method flow for constructing a valid item set for recommendation matching is as follows:

[0045] Step 11: Obtain the viewing log record data of the IPTV user, extract the event data corresponding to the user behavior state including the user ID, time, and channel ID field information, and establish the main feature set of the event data;

[0046] Step 12: Take the IPTV channel ID as the user's behavior item set, form the user's item sequence according to the order in which the behavior is generated, that is, the time sequence in which the user watches the channel, and construct a combination of the target event and the user-event interaction sequence. The target event is represented by its main features as the valid item set, and the user-event interaction is represented by a set of event features containing time codes;

[0047] Step 13: Expand the single mode to multi-mode from the user's item sequence, extract the scalable multi-mode item set to construct an effective feature set for recommendation matching, and take the order information of the user's viewing channels into consideration in feature extraction, considering not only the user features formed by a single channel, but also the connection between channels;

[0048] Step 14: The constructed recommended matching valid item set is transmitted to the IPTV data mining and analysis terminal, and the mining and analysis terminal performs analysis and mining based on the received three types of data including user data, program data, and user viewing log.

[0049] Step 2: Mining users’ implicit feedback in IPTV based on the constructed recommendation matching valid item set;

[0050] In an embodiment of the present invention, a method for mining implicit feedback of a user in IPTV includes:

[0051] Step 21: Establish a user interest mining model based on the received user data, program data, and user viewing log data, quantify the user viewing behavior, reflect the user interest through the user viewing behavior, and mine the viewing behavior that can stimulate the user;

[0052] Step 22: Synchronously extract the viewing time corresponding to the user's viewing record event, use the viewing time analysis to determine the user's interest in the current video, propose a user classification algorithm based on the user's viewing time to classify the user, and analyze the user's implicit viewing habits.

[0053] Step 23: Linking user interests and program types through IPTV program type labels, and building a label-based user viewing time sequence matrix based on the proportion of user viewing time in a specified time period, and extracting user implicit interest features using a clustering method based on a Gaussian mixture model;

[0054] Step 24: Utilizing the temporal characteristics of the viewing records of users in the IPTV system, the multi-modal item set is used as the identification input of the user interest mining model. Since IPTV is user-oriented, the viewing behavior of a user ID may reflect the interests of different individuals. The user's implicit interests are extracted using a recurrent neural network. An attention mechanism is added to the recurrent neural network model to distinguish the impact of different historical interest states on the current user's viewing decision.

[0055] Step 25: Label other users with similar target interests to the current user with corresponding tags and classify them into a total user implicit feedback set;

[0056] Step 26: Describe the implicit feedback characteristics of users in IPTV through annotated tags, use coarse-grained tags to represent the types of videos that users are interested in hidden in the IPTV system, and use fine-grained tags to represent the implicit connection between the types of videos that users are interested in and the system recommendations based on the tags.

[0057] Step 3: Using the user's viewing time and the total video time in the implicit connection, calculate the importance ranking of the implicit factors that affect the recommendation of video programs in the IPTV system;

[0058] For example, in an IPTV system, the total duration of a video program is t0, and the duration of a user watching the video program is t1. The duration of a user watching the video program t1 and the total duration of the video program t0 are used to calculate the ratio of the user watching time to the total duration of the video, that is, t 比值 =t1 / t0, calculate and summarize the ratios of all types of video programs watched by the user to obtain the ratio set ∑t of all types of video programs watched by the user i , t i ={t 比值1 , t 比值2 , t 比值3 ,…,t 比值n}, through the ratio set ∑t i Calculate the average proportion of all video programs watched by users As a result, by calculating t 比值 Ratio to average The logarithm k of the user's viewing interest is used to judge whether it exceeds the average value, and the attribute interval of k is set to k∈(-1,1), and When k is a positive number, it is judged that the user's viewing completion of the current video program is higher than the average, which means that the user has a higher interest and preference for the current video program. When k is a negative number, it is judged that the user's viewing completion of the current video program is lower than the average, which means that the user has a lower interest and preference for the current video program. The output is sorted from small to large according to the attribute interval of the k value.

[0059] Step 4: Calculate the selection rate of the user's interested video program type based on the mined user implicit feedback data;

[0060] In the embodiment of the present invention, the time period is set according to the mined implicit feedback data of the user, and the unit of the time period is set to hours, each hour represents a time period, and in the IPTV family group user recommendation scenario, the time period with the most appearances in all the user's viewing records is defined as the main time period T0, and k is the time period T0 obtained by calculating t 比值 Ratio to average The logarithm of is used to reflect whether the user's viewing interest exceeds the average value. For family users, the change in interest is mainly caused by the switching of members. The selection rate of the user in the specified time period is established based on the similarity of the viewing time periods of the members. That is, the user's selection rate is Q. By measuring the probability of switching behavior of family group users based on viewing time, the impact of member switching on users' interest preferences for current video programs is captured.

[0061] Step 5: Based on the user's existing historical viewing behavior data, according to the behavior records of users with similar interests, expand the recommendation of video programs that the user has never watched but meet the user's implicit needs.

[0062] In an embodiment of the present invention, automatic clustering based on user behavior statistics is performed. When different types of video programs are selected and clicked by multiple users at the same time, user interest features and user classification based on viewing time statistics are used as the main influencing features of user selection rate. The video programs are divided into implicitly associated categories, namely, light viewing, normal viewing, high viewing, and heavy viewing. The granularity of the classification is controlled according to the implicitly associated categories. The more categories there are, the finer the granularity will be. The weight of each video program belonging to each category is calculated. When users who like a certain category will choose one of the video programs, then the weight of the video program in this category is higher. The video programs in the set that have not been subscribed by the target user are sorted in descending order according to the subscription frequency and selection rate. According to this weight value, video media assets that meet the user's interests but have never been selected by the user are recommended to the user, forming a surprising content recommendation list of different regions based on the user's implicit feedback.

[0063] Embodiment 2: Embodiment 2 of the present invention provides an interactive network television service system. Figure 2 A schematic diagram of the module composition of an interactive network television service system provided in the second embodiment of the present invention is shown in FIG. Figure 2 As shown, the system includes:

[0064] The effective item set construction module is used to construct the effective item set for recommendation matching by using the behavior data of IPTV users;

[0065] Implicit feedback mining and analysis module, used to mine users' implicit feedback in IPTV based on the constructed recommendation matching valid item set;

[0066] The implicit feedback recommendation module is used to expand the recommendation of video programs that the user has never watched but meets the user's implicit needs.

[0067] In some embodiments of the present invention, the valid itemset construction module includes:

[0068] The behavior data module uses the IPTV channel ID as the user's behavior item set to obtain behavior data in the order in which the behavior occurs;

[0069] A feature set module is used to establish a main feature set of event data according to the time data corresponding to the behavior state;

[0070] The multimodal itemset construction module is used to extract an effective feature set for scalable multimodal itemset construction recommendation matching.

[0071] In some embodiments of the present invention, the implicit feedback mining and analysis module includes:

[0072] An interest mining model module is used to establish a user interest mining model based on received user data, program data, and user viewing log data;

[0073] Implicit viewing habit analysis module, used to classify users according to their viewing time by proposing user classification algorithms and analyzing their implicit viewing habits;

[0074] An implicit interest feature extraction module is used to extract the user's implicit interest features using a clustering method based on a Gaussian mixture model;

[0075] The implicit factor ranking module is used to calculate the importance ranking of implicit factors that affect the recommendation of video programs in the IPTV system.

[0076] In some embodiments of the present invention, the implicit feedback recommendation module includes:

[0077] A selection rate calculation module, used to calculate the selection rate of the video program type that the user is interested in;

[0078] An implicit association classification module is used to classify video programs into implicitly associated categories;

[0079] The extended recommendation module is used to form a surprising content recommendation list of different regions based on user implicit feedback.

[0080] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0081] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An interactive network television service method, characterized in that: The method comprises the following steps: Step 1: Use IPTV user behavior data to construct a valid item set for recommendation matching; Step 2: Mining users’ implicit feedback in IPTV based on the constructed recommendation matching valid item set; Step 3: Using the user's viewing time and the total video time in the implicit connection, calculate the importance ranking of the implicit factors that affect the recommendation of video programs in the IPTV system; Step 4: Calculate the selection rate of the user's interested video program type based on the mined user implicit feedback data; Step 5: Based on the user's existing historical viewing behavior data, according to the behavior records of users with similar interests, expand the recommendation of video programs that the user has never watched but meet the user's implicit needs; The method flow steps of constructing a valid item set for recommendation matching include: Step 11: Obtain the viewing log record data of the IPTV user, extract the event data corresponding to the user behavior state including the user ID, time, and channel ID field information, and establish the main feature set of the event data; Step 12: Take the IPTV channel ID as the user's behavior item set, form the user's item sequence according to the order in which the behavior is generated, that is, the time sequence in which the user watches the channel, and construct a combination of the target event and the user-event interaction sequence. The target event is represented by its main features as the valid item set, and the user-event interaction is represented by a set of event features containing time codes; Step 13: Expand the single mode to multi-mode from the user's item sequence, extract the scalable multi-mode item set to construct an effective feature set for recommendation matching, and take the order information of the user's viewing channels into consideration in feature extraction, considering not only the user features formed by a single channel, but also the connection between channels; Step 14: The constructed recommended matching valid item set is transmitted to the IPTV data mining and analysis terminal, and the mining and analysis terminal performs analysis and mining based on the received three types of data including user data, program data, and user viewing log; The method steps for mining implicit feedback of users in IPTV include: Step 21: Establish a user interest mining model based on the received user data, program data, and user viewing log data, quantify the user viewing behavior, reflect the user interest through the user viewing behavior, and mine the viewing behavior that can stimulate the user; Step 22: Synchronously extract the viewing time corresponding to the user's viewing record event, use the viewing time analysis to determine the user's interest in the current video, propose a user classification algorithm based on the user's viewing time to classify the user, and analyze the user's implicit viewing habits. Step 23: Linking user interests and program types through IPTV program type labels, and building a label-based user viewing time sequence matrix based on the proportion of user viewing time in a specified time period, and extracting user implicit interest features using a clustering method based on a Gaussian mixture model; Step 24: Utilizing the temporal characteristics of the viewing records of users in the IPTV system, the multi-modal item set is used as the identification input of the user interest mining model. Since IPTV is user-oriented, the viewing behavior of a user ID may reflect the interests of different individuals. The user's implicit interests are extracted using a recurrent neural network. An attention mechanism is added to the recurrent neural network model to distinguish the impact of different historical interest states on the current user's viewing decision. Step 25: Label other users with similar target interests to the current user with corresponding tags and classify them into a total user implicit feedback set; Step 26: Describe the implicit feedback characteristics of users in IPTV through the annotated tags, use coarse-grained tags to represent the types of videos that users are interested in hidden in the IPTV system, and use fine-grained tags to represent the implicit connection between the IPTV system and the system recommendation matching of the types of videos that users are interested in based on the tags; The step of calculating the importance ranking includes: In the IPTV system, the total duration of the video program is t0, and the duration of the user watching the video program is t1. The duration t1 of the user watching the video program and the total duration t0 of the video program are used to calculate the ratio of the user watching time to the total duration of the video, that is, t 比值 =t1 / t0, calculate and summarize the ratios of all types of video programs watched by the user to obtain the ratio set ∑t of all types of video programs watched by the user i , t i ={t 比值1 , t 比值2 , t 比值3 ,…,t 比值n }, through the ratio set ∑t i Calculate the average proportion of all video programs watched by users As a result, by calculating t 比值 Ratio to average The logarithm k of the user's viewing interest is used to judge whether it exceeds the average value, and the attribute interval of k is set to k∈(-1,1), and When k is a positive number, it is judged that the user's viewing completion of the current video program is higher than the average value, which means that the user has a higher interest and preference for the current video program. When k is a negative number, it is judged that the user's viewing completion of the current video program is lower than the average value, which means that the user has a lower interest and preference for the current video program. The output is sorted from small to large according to the attribute interval of the k value; The step of calculating the selection rate of the user for the video program type of interest comprises: The time period is set according to the mined implicit feedback data of users, and the unit of the time period is set to hours, each hour represents a time period. In the IPTV family group user recommendation scenario, the time period with the most appearances in all the user's viewing records is defined as the main time period T0, and k is the time period T0 obtained by calculating t 比值 Ratio to average The logarithm of is used to reflect whether the user's viewing interest exceeds the average value. For family users, the change in interest is mainly caused by the switching of members. The selection rate of the user in the specified time period is established based on the similarity of the viewing time periods of the members. That is, the user's selection rate is Q. By measuring the probability of switching behavior of family group users based on viewing time, the impact of member switching on users' interest preference for current video programs is captured; The steps of extending the recommendation include: Automatic clustering based on user behavior statistics. When different types of video programs are selected and clicked by multiple users at the same time, user interest characteristics and user classification based on viewing time statistics are used as the main influencing features of user selection rate. Video programs are divided into implicitly associated categories, namely light viewing, normal viewing, high viewing, and heavy viewing. The granularity of classification is controlled according to the implicitly associated categories. The more categories there are, the finer the granularity will be. The weight of each video program belonging to each category is calculated. When users who like a certain category will choose one of the video programs, then the weight of the video program in this category is higher. The video programs in the set that the target user has not subscribed to are sorted in descending order according to the subscription frequency and selection rate. According to this weight value, video media assets that meet the user's interests but have never been selected by the user are recommended to the user, forming a surprising content recommendation list of different regions based on the user's implicit feedback.

2. An interactive network television service system for executing the interactive network television service method according to claim 1, characterized in that: The system comprises: The effective item set construction module is used to construct the effective item set for recommendation matching by using the behavior data of IPTV users; Implicit feedback mining and analysis module, used to mine users' implicit feedback in IPTV based on the constructed recommendation matching valid item set; The implicit feedback recommendation module is used to expand the recommendation of video programs that the user has never watched but meets the user's implicit needs; The valid item set construction module includes: The behavior data module uses the IPTV channel ID as the user's behavior item set to obtain behavior data in the order in which the behavior occurs; A feature set module is used to establish a main feature set of event data according to the time data corresponding to the behavior state; Multimodal itemset construction module, used to extract effective feature sets for scalable multimodal itemset construction recommendation matching; The implicit feedback mining and analysis module includes: An interest mining model module is used to establish a user interest mining model based on received user data, program data, and user viewing log data; Implicit viewing habit analysis module, used to classify users according to their viewing time by proposing user classification algorithms and analyzing their implicit viewing habits; An implicit interest feature extraction module is used to extract the user's implicit interest features using a clustering method based on a Gaussian mixture model; An implicit factor ranking module is used to calculate the importance ranking of implicit factors that affect the recommendation of video programs in the IPTV system; The implicit feedback recommendation module comprises: A selection rate calculation module, used to calculate the selection rate of the video program type that the user is interested in; An implicit association classification module is used to classify video programs into implicitly associated categories; The extended recommendation module is used to form a surprising content recommendation list of different regions based on user implicit feedback.

Images