KR101054088B1 - How to automatically recommend customized IP programs using collaborative filtering - Google Patents

Abstract

The present invention relates to a method for recommending broadcast program information to a viewer's terminal in a TV or IP TV service. The present invention relates to a method for collaborative filtering from a viewer group's viewing program that matches the viewer's taste or has a viewer's taste in the terminal. It extracts the used programs and sorts them in order of high preference, and automatically recommends several programs with high preference. In addition, by inferring viewers' viewing order of broadcasting programs and presenting customized personal broadcasting program schedules, the burden of searching for and selecting broadcasting programs can be greatly reduced, and the efficiency of broadcasting services is further improved by providing personalized broadcasting programs. There is an advantage.

Program recommendations, personalized program schedules, collaborative filtering, sequential pattern mining, similar viewer clustering

Description

  How to automatically recommend customized IP programs using collaborative filtering {A METHOD OF USER-CUSTOMIZED IP TV PROGRAM RECOMMENDATION USING COLLABORATION FILTERING}

The present invention relates to a technology for recommending a user's favorite program in IP TV, and in particular, recommends a user-tailored broadcast program through collaborative filtering of program viewing data of other users similar to the user's viewing preference, The present invention relates to a method for automatically recommending a customized ID program using collaborative filtering, which is suitable for analyzing a user's viewing pattern and providing personalized program information.

In general, when purchasing an item on an Internet purchase site, a collaborative filtering on items purchased by the recommended buyer and similar taste users is recommended to the user to facilitate the purchase of the item and to sell the item. Augmentation techniques are used.

Traditionally, collaborative filtering techniques group similar users with similar purchase preferences based on the purchase history of each individual user, and then resemble similar preferences for purchases that the active user did not purchase. It is a technique that sorts and recommends items purchased by users.

In general, collaborative filtering uses preference values that an individual expresses explicitly for the purchased item, such as a star rating or a comment. However, in the current Internet Protocol Television (IPTV) environment, it is very difficult to require a user to explicitly enter a preference for the program content that is viewed, and it is cumbersome for the user to easily collect data on a reliable preference. Not.

This is related to the fact that the user of the IP TV terminal (hereinafter referred to as the 'viewer') does not provide the convenience of the interface for input yet, which is a obstacle to applying the IP TV service to the collaborative filtering. In addition, most methods have a limitation in that they can not automatically make recommendations for program contents that have not been watched by the viewers by making personalized program recommendations using only the user's own viewing program data.

Accordingly, an object of the present invention is to group IP TV viewers into one or more groups of viewers with similar viewing preferences and to collaborate with program viewers 'historical data of recommended viewers and similar viewer groups having similar viewing preferences and program viewers' historical data of recommended viewers. After selecting the candidate recommendation program content by filtering, the recommendation order is used in order of priority and recommendation order.

Another object of the present invention is to consider the fact that viewers have a broadcast program viewing order pattern by day / time zone when a viewer watches a broadcast program, and according to the sequence pattern mining technique of the program viewing history data of the viewer to organize a personalized broadcast program viewing order To provide information.

The objects of the present invention are not limited to the above-mentioned objects. Other objects and advantages of the invention will be more clearly understood by the following description.

A first aspect of the present invention for achieving the above object is, by considering one or more of the program preferences, genre preferences and channel preferences for the broadcast program viewing history data of multiple IP TV viewers to the multiple IP TV group of one or more similar viewers Grouping viewers and providing similar viewer groups with a way to find a group of viewers most similar to the recommended audience.

The second aspect of the present invention implicitly calculates the preferences for the broadcast program in consideration of the length and the number of viewers watching the broadcast program instead of the explicit method of requesting input as a rating indicating the viewer's interest in the broadcast program. Based on this, a candidate recommendation broadcast is performed by performing collaborative filtering on program viewing history data of a similar viewer group and similar viewing viewer's program viewing history data using a method of extracting interest program content. The present invention provides a method of selecting program content to generate a recommendation candidate program.

According to a third aspect of the present invention, an IP TV viewer can easily access a recommended program by determining a recommendation priority by recommending a ranking ordering model for the selected recommendation candidate programs and recommending one or more programs. Is to provide a way to do this.

According to a fourth aspect of the present invention, an IP TV viewer can change a channel while viewing a program by recommending a personalized program by using a sequential pattern mining technique for a group of similar taste viewers and an IP TV viewer. It provides a way to make it easier to access preferred programs by reducing the hassle of searching for them.

A fifth aspect of the invention includes the steps of: (a) grouping a plurality of viewers into one or more similar viewer group units; (b) confirming that the viewer logs in to the IP TV terminal; (c) selecting one or more most similar viewer groups among the similar viewer groups for viewers logged in to the IP TV terminal; (d) detecting recommended candidate programs using collaborative filtering on the program viewing history data of the selected similar viewer groups and the history data of the viewer logged in to the IP TV terminal; (e) determining a recommendation rank through ranking alignment with respect to the detected recommendation candidate programs and recommending one or more recommendation candidate programs having a high recommendation rank as the final recommendation program to the IP TV terminal. It is done.

In the step (a), when grouping a plurality of viewers in units of one or more similar viewer groups, fuzzy clustering may be used to allow one viewer to belong to one or more similar viewer groups.

In the step (d), the candidate candidate program is characterized in that the viewer watched more than 10% of the total program amount.

A sixth aspect of the invention includes the steps of: (a) grouping a plurality of viewers into one or more similar viewer group units; (b) confirming that the viewer logs in to the IP TV terminal; (c) selecting one or more most similar viewer groups among the similar viewer groups for the viewers logged in to the IP TV terminal; (d ') constructing recommended programs in sequential order using sequential pattern mining on the program viewing history data of the selected similar viewer groups and the history data of the viewer logged in to the IP TV terminal; (e ') presenting the programs arranged in the chronological order in the form of a program schedule.

In the step (d '), it is characterized by determining in consideration of the program preferences of the viewer when configuring the recommended programs in a time sequence.

When configuring the recommended programs in the time order in the step (d '), the frequent viewing program is extracted from the program viewing history data and configured.

A seventh aspect of the present invention provides a recording medium on which a program is recorded for executing a program automatic recommendation in a personalized program automatic recommendation.

The present invention collaboratively filters program viewing history data of a similar viewer group and viewer viewing preferences similar to the viewers who are recommended to watch, and selects candidate candidate broadcast program contents by jointly filtering the program viewing history data of the viewers who are recommended. By recommending the broadcast program content to the IP TV terminal in the order of high recommendation ranking, the viewer can easily select a program and access to the corresponding program content is very easy.

In addition, according to the present invention, in view of the broadcast program viewing order pattern by day / time zone when a viewer watches a broadcast program, the order of viewing a personalized broadcast program by applying a sequential pattern mining technique to the viewer's program viewing history data. By providing the programming, the viewer can easily access the broadcast program, thereby improving the program viewing convenience.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the following embodiments of the present invention may be modified into various other forms, and the scope of the present invention is not limited to the embodiments described below. The embodiments of the present invention are provided to more completely explain the present invention to those skilled in the art.

의 프로그램

에 대한 프로그램 선호도

는 다음의 [수학식]과 같이 정의한다.In the present invention, the viewer's preference for the content of the IP TV program is defined as the number of viewing interests compared to the total number of IP TV program broadcasts. At this time, in order to be considered as a program of interest viewing, it is based on the fact that the actual viewing length is 10% or more of the total compared to the program airing length. This is to increase reliability by excluding programs watched for a while while changing channels from the program of interest. In addition, series programs that are continuously aired under one title, such as news and drama, are regarded as one program. In other words, viewers

Program of

Program affinity for

Is defined as the following [Equation].

는 시청자

의 프로그램

의 관심 시청 여부를 0 또는 1로 나타내며,

는 시청자

의 개별 프로그램

로 이루어진 시리즈물 전체 프로그램

의 총 시청 횟수를 나타낸다.

는 시리즈물 전체 프로그램

의 총 방송 횟수를 나타낸다. 그리고

는 프로그램

의 방영 길이를 나타내고

는 시청자

가 프로그램을 실제 시청한 길이를 나타낸다. here,

Viewers

Program of

Indicates 0 or 1 when watching

Viewers

Individual program of

The complete series of series

Indicates the total number of views.

The series full program

Indicates the total number of broadcasts. And

Program

Indicates the airing length of

Viewers

Represents the length of the actual viewing of the program.

라 할 때, 장르

에 대한 시청자

의 선호도

를 다음의 [수학식 4]와 같이 정의한다. The preference is calculated by adding a preference value of the genre to which the program belongs using the preference for the program. The genre preference value is normalized by dividing the value of the preference belonging to all genres. The total number of genres

When I say genre

Viewers for

Affinity

Is defined as in Equation 4 below.

라 할 때, 시청자

의 채널

에 대한 선호도

는 다음의 [수학식 5]와 같이 정의한다. In the same way as the genre preference is calculated, the total number of channels

When we say viewers

'S channel

Affinity for

Is defined as in Equation 5 below.

When clustering a large number of viewers, a reasonable sized unit considering the efficiency of the calculation is needed. By using domain information or location information, viewers can be divided into a certain number of basic units. This unit can be determined by the size of the similar taste viewer group. A widely used technique for clustering is C-means clustering or dynamic fuzzy clustering technique that may have elements belonging to members of multiple clusters.

와 채널에 대한 선호도

는 다음의 [수학식 6],[수학식 7]과 같이 정의한다. For calculating the preference of each cluster, the preference for the genre of each cluster may be calculated as the sum of the preferences for the genres of viewers belonging to the cluster. Then, it is divided by the number of viewers in the cluster and normalized. In the same way, the preferences for the channel of the cluster and the viewing time are calculated. Affinity for the genre of each cluster

And preferences for channels

Is defined as [Equation 6], [Equation 7] below.

A viewer who logs in via an IP TV viewer terminal becomes an active viewer and a similar taste viewer group with membership is determined. In this similar taste viewer group, a process of extracting viewers with similar viewing tastes is specifically performed. Pearson Correlation Coefficient (PCC) may be used as a method of measuring similarity between viewers of an active viewer and a similar taste viewer group. The PCC uses the normalized value as the base vector by subtracting the viewer's preference value from the average preference value.

의 선호도 프로파일

로 정의 한다.

에 대해 활동시청자의 선호도 값과 추출된 유사 취향 시청자 그룹의 각 멤버간의 선호도에 대한 유사도를 계산한다. 이 과정을 통해, 활동 시청자와 선호도가 유사한 시청자를 추출한다. 활동 시청자

와 시청자

간의 유사도

는 다음의 [수학식 8]과 같이 측정된다.Viewers of the program content watched by the activity viewer

Affinity Profile

It is defined as

For, we calculate the similarity of the preference value between the activity viewer and each member of the extracted similar taste viewer group. Through this process, viewers with similar preferences as active viewers are extracted. Activity viewer

And viewers

Similarity between

Is measured as in Equation 8 below.

은 활동시청자가 관심을 보인 시청 프로그램들을 나타내며,

은 시청자

의 프로그램

에 대한 선호도 값이고,

은 시청자

의 평균 시청 선호도이다. 그리고,

는 활동 시청자

가 시청한 관심 프로그램 집합을 나타낸다.here,

Indicates the viewing programs of which the viewer was interested,

Silver viewers

Program of

Is an affinity value for,

Silver viewers

Average viewing preference. And,

Are active viewers

Represents a set of programs of interest.

After selecting viewers with similar viewing preferences to active viewers, their favorite viewing programs are selected as candidate programs for recommendation. When recommending a real-time broadcast program, only programs accessible through EPG information can be selected. In the case of including program content in the form of VOD such as IP TV, the EPG can be used as information on what the latest program is.

와 질의어

의 관계를 추천 후보 프로그램

와 활동 시청자

와의 관계로 재해석 한다. 그리고, 활동 시청자

를 활동시청자가 선호하는 프로그램의 프로파일

로 대체함으로써, 추천 후보 프로그램

와 활동 시청자 시청 선호 프로그램

간의 관계는 다음 [수학식 9]로 정의된다.First, the search result candidate document

And query

Nomination program

And activity viewers

Interpret in relation to And activity viewer

Profile of the program preferred by the viewer

By substituting the referral candidate program

And activity viewers prefer watching

The relationship between is defined by Equation 9 below.

here,

은

프로그램들의 전체 방영 횟수를, 그리고

은

프로그램의 방영 횟수를 나타낸다.

는 앞에서 설명된 바와 같이 추출된 후보 프로그램들을 나타내 며,

는

에 속하는 활동 시청자의 선호 프로그램을 의미한다.

는 유사 취향 시청자들

의 전체 프로그램

시청 횟수 대비,

와

두 프로그램을 시청한 시청자들의 시청횟수의 합의 비율을 나타낸다. 이 때, 유사 취향 시청자들의 명수를

라 하면, 프로그램을 시청한 횟수의 비율은 다음의 [수학식 10]으로 정의된다.

silver

The total number of shows aired, and

silver

Shows the number of times the program aired.

Represents candidate programs extracted as described above.

Is

It means a favorite program of activity viewers belonging to.

Similar taste viewers

Full program of

Of viewcounts,

Wow

It shows the ratio of the sum of the number of viewers who watched two programs. At this time, the number of viewers

In this case, the ratio of the number of times the program is viewed is defined by Equation 10 below.

는 활동 시청자의 프로그램

시청 횟수 대비,

프로그램을 시청한 횟수의 비율을 의미한다.

Program of activity viewers

Of watched times,

The ratio of the number of times a program is watched.

는 두 아이템

와

간의 유사도에 대한 가중치로, 시청자-아이템(User-Item) 매트릭스에서 VCC(Vector Cosine Correlation) 방법으로 다음의 [수학식 11]과 같이 측정된다.

Two items

Wow

The weight of the similarity between the two is measured by the Vector Cosine Correlation (VCC) method in the viewer-item matrix as shown in Equation 11 below.

Meanwhile, referring to FIG. 1, a process of an embodiment of a method for automatically recommending a viewer customized ATV program using collaborative filtering according to the present invention will be described below.

Normally, a plurality of viewers are grouped into one or more similar viewer group units by referring to the IP TV viewer database 1 (SA1).

If it is determined that the viewer logs in by checking whether the viewer logs in to the terminal, one or more most similar viewer groups among the similar viewer groups are selected for the logged in viewer (SA2, SA3).

The recommended candidate programs are selected using the collaborative filtering on the program viewing history data of the selected similar viewer groups and the history data of the viewers logged in to the IP TV terminal. (SA4, SA5)

Subsequently, the ranking of the selected recommendation candidate programs is determined by rank ordering, and one or more recommendation candidate programs having a high recommendation ranking are recommended as the final recommendation program to the IP TV terminal. (SA6, SA7)

Meanwhile, referring to FIG. 2, a process of another embodiment of the method for automatically recommending a viewer customized TV program using collaborative filtering according to the present invention will be described.

Normally, a plurality of viewers are grouped into one or more similar viewer group units by referring to the IP TV viewer database 1 (SB1).

If it is determined that the viewer logs in by checking whether the viewer logs in to the terminal, one or more most similar viewer groups among the similar viewer groups are selected for the logged in viewer (SB2, SB3).

The recommended candidate programs are arranged in chronological order using sequential pattern mining on the program viewing history data of the selected similar viewer groups and the history data of the viewers logged in to the IP TV terminal. (SB4, SB5)

Thereafter, the programs arranged in chronological order are presented (recommended) to the IP TV terminal in the form of a program schedule. (SB6)

Meanwhile, the method for automatically recommending a personalized program according to an exemplary embodiment of the present invention may also be implemented as computer readable codes on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored.

For example, the computer-readable recording medium may be a ROM, a RAM, a CD-ROM, a magnetic tape, a hard disk, a floppy disk, a removable storage device, a nonvolatile memory (Flash memory). Optical data storage, and the like, and also implemented in the form of a carrier wave (eg, transmission over the Internet).

The computer readable recording medium can also be distributed over computer systems connected over a computer network so that the computer readable code is stored and executed in a distributed fashion.

Although a preferred embodiment of the method for automatically recommending a personalized program according to the present invention has been described above, the present invention is not limited thereto and is variously modified within the scope of the claims and the detailed description of the invention and the accompanying drawings. It is possible to implement and this also belongs to the present invention.

Although the preferred embodiment of the present invention has been described in detail above, the scope of the present invention is not limited thereto, and may be implemented in various embodiments based on the basic concept of the present invention defined in the following claims. Such embodiments are also within the scope of the present invention.

1 is a flowchart of an embodiment of a method for automatically recommending a viewer customized IP program using collaborative filtering according to the present invention.

Figure 2 is a flow diagram for another embodiment of a method for automatically recommending viewer customized IP program using collaborative filtering according to the present invention.

*** Description of the symbols for the main parts of the drawings ***

1: IP TV viewer database

Claims (8)

In a method for recommending a viewer's favorite program to an IP TV terminal, (a) grouping a plurality of viewers into one or more similar viewer group units using fuzzy clustering to enable one viewer to belong to one or more similar viewer groups; (b) confirming that the viewer logs in to the IP TV terminal; (c) one of the similar viewer groups for the viewer logged into the IP TV terminal. Selecting one or more most similar viewer groups; (d) detecting recommended candidate programs using collaborative filtering on the program viewing history data of the selected similar viewer groups and the history data of the viewer logged in to the IP TV terminal; (e) determining a recommendation rank through ranking alignment with respect to the detected recommendation candidate programs and recommending one or more recommendation candidate programs having a high recommendation rank as the final recommendation program to the IP TV terminal. How to automatically recommend customized TV programs using collaborative filtering. delete The active viewer according to claim 1, wherein in order to select a similar viewer group in the step (c),

And viewers

Similarity between

Auto-recommended method for the user customized ID program using collaborative filtering, characterized in that obtaining the.

here,

Is the viewing program that interests viewers,

Silver viewers

Program of

Affinity value for,

Is the viewer's average viewing preferences,

Are active viewers

Set of shows watched by. The method of claim 1, wherein the recommendation candidate program in step (d) is a program that a viewer watches at least 10% of the total program amount. In a method for recommending a viewer's favorite program to an IP TV terminal, (a) grouping a plurality of viewers into one or more similar viewer group units; (b) confirming that the viewer logs in to the IP TV terminal; (c) selecting one or more most similar viewer groups among the similar viewer groups for the viewers logged in to the IP TV terminal; (d ') sequential pattern mining is performed on the program viewing history data of the selected similar viewer groups and the history data of the viewer logged in to the IP TV terminal, and the recommended programs determined in consideration of the program preference of the viewer in chronological order. Constructing; (e ') presenting the programs arranged in the order of time in the form of a program schedule, the method of automatically organizing a personalized program using sequential pattern mining, comprising: a. delete The method of claim 5, wherein when the recommendation programs are configured in a time sequence in step (d '), a personalized program using sequential pattern mining is automatically extracted from the program viewing history data. How to organize. Claim 8 was abandoned when the registration fee was paid. A computer-readable recording medium having recorded thereon a program capable of executing the method of claim 1 or 3.

Images