KR20150101341A - Apparatus and method for recommending movie based on distributed fuzzy association rules mining - Google Patents

Abstract

Disclosed are a device and a method for recommending a movie based on a distributed fuzzy association rule mining. The present invention acquires first rating data comprising a movie rating, converts the acquired first rating data to second rating data using a fuzzy membership function, and generates a list of an associated movie by applying the converted second rating data to a fuzzy association rule mining. Recommending a movie associated with a user who is to receive recommendation is available by recommending a movie in the list of the associated movie in descending order by using the generated list of the associated movie.

Description

[0001] APPARATUS AND METHOD FOR RECOMMENDING MOVIE BASED ON DISTRIBUTED FUZZY ASSOCIATION RULES MINING [0002]

The present invention relates to a movie recommendation apparatus and method based on distributed fuzzy association rule mining that converts a movie rating information of a user into linguistic information and recommends a movie through association of linguistic information, To a movie recommendation apparatus and method based on distributed fuzzy association rule mining capable of generating a list of related movies acquired by applying to a fuzzy association rule mining and recommending movies to a recommendation target user using the generated list.

A movie recommendation through a general recommendation algorithm is recommended in association rule mining based on a user's movie purchase history in the form of a movie viewed by a person who viewed the movie. Alternatively, it is possible to calculate the similarity between movies and extract 'movies similar to this movie'.

However, it is difficult to find an example of applying fuzzy association rule mining to a movie recommendation domain because it is difficult to calculate the association rule using a conventional method when the rating log of the movie is large and the number of users is large.

Therefore, the user's rating information about the movie is converted into linguistic information, the association relation of the user's movie rating information is acquired through the distributed fuzzy association rules mining based on the linguistic information, There is an urgent need for movie recommendation technology based on distributed fuzzy associative rule mining that can recommend a suitable movie.

Korean Patent Laid-Open No. 10-2013-0009360A, July 30, 2013 (name: method and system for providing movie recommendation service)

An object of the present invention is to recommend a movie suitable for a user's preference propensity by replacing the rating information left by the users with the linguistic evaluation information and recommending the movie related to the replaced evaluation information.

Another object of the present invention is to provide a reliable movie recommendation function to users by efficiently processing movie rating information of a large capacity using a data processing method suitable for a distributed framework.

According to an aspect of the present invention, there is provided a movie recommendation apparatus including a data acquiring unit acquiring first rating data including a rating for a movie, converting the acquired first rating data to second rating data, An association list generation unit for applying the second rating data to the fuzzy association rule mining to generate an associated movie list, and a movie recommendation unit for recommending the movie using the associated movie list to the recommendation target user.

At this time, the association list generation unit assigns ratings to at least one of a fuzzy membership function including a triangular membership function, a trapezoidal membership function, and a Gaussian membership function, And converting the language label according to the obtained fuzzy membership value into the second score data by replacing the score with the score.

At this time, the association list generation unit may generate at least one of fuzzy reliability and fuzzy correlation using fuzzy association rule mining, and generate a related movie list based on at least one of the generated fuzzy reliability and fuzzy correlation.

In this case, the association list generation unit generates an association combination generation unit for generating a film-by-film association combination by combining the converted second rating data according to the fuzzy association rule, and a rating history for each movie, And a fuzzy support calculation unit for calculating fuzzy support for each movie using the generated rating history of each movie.

In this case, the fuzzy support calculation unit can calculate the fuzzy support per film by using the reference value obtained by normalizing the fuzzy membership value.

At this time, the association list generation unit calculates at least two of the fuzzy membership functions so as to calculate the association fuzzy support for the movie-related association, and the fuzzy membership reliability is calculated using at least one of the fuzzy support of the movie and the calculated association- Can be calculated.

In this case, the association list generation unit may calculate the fuzzy correlation using one or more of the fuzzy support per fuze, the fuzzy reliability, and the square value of the fuzzy support per movie.

At this time, the movie recommendation unit determines the ranking of the associated movie list generated according to the predetermined importance level, and recommends the movie in the order of the determined associated high movie list.

At this time, the associated movie list may include one or more of the title, genre, director, country, year and image of the movie.

The movie recommendation method according to the present invention includes the steps of acquiring input data including a rating for a movie, converting the acquired input data to second rating data, and applying the converted second rating data to fuzzy association rule mining Creating a list of related movies, and recommending movies using a list of associated movies created for the user to be recommended.

In this case, the step of generating the associated movie list may include assigning a rating to at least one of a fuzzy membership function including a triangular membership function, a trapezoidal membership function, and a Gaussian membership function And converting the language label according to the obtained fuzzy belonging value into the second score data by replacing the score with the score.

The generating of the associated movie list may include generating at least one of fuzzy reliability and fuzzy correlation using association rule mining, generating a list of associated movies based on at least one of the generated fuzzy reliability and fuzzy correlation, .

At this time, the step of generating the associated movie list may include generating at least one of fuzzy reliability and fuzzy correlation using fuzzy association rule mining, and generating a related movie list based on one of the generated fuzzy reliability and fuzzy correlation .

In this case, the step of generating the associated movie list may include combining at least two of the fuzzy membership functions to calculate the association fuzzy support for the movie-related association, and calculating at least one of the fuzzy support for the movie and the associated combination fuzzy support And calculating the fuzzy reliability using the fuzzy reliability.

In this case, the step of generating the associated movie list may include calculating the fuzzy correlation using at least one of the values of the fuzzy support of the movie, the fuzzy reliability, and the square value of the fuzzy support of the movie.

At this time, the step of recommending a movie may include a step of determining a ranking of the associated movie list generated according to a predetermined importance, and the movie may be recommended in the order of the determined associated high movie list.

According to the present invention, rating information about a movie is acquired from a plurality of users, and a list of related movies is extracted using the acquired rating information. By using the rating of a user to recommend a movie, You can recommend a movie.

In addition, according to the present invention, since the score of the movie is replaced with the linguistic information to derive the association, various recommended movie lists can be generated and provided based on various directions of the linguistic information.

1 is a block diagram showing a movie recommendation apparatus according to an embodiment of the present invention.
FIG. 2 is a block diagram illustrating an association list generation unit of the movie recommendation apparatus of FIG. 1. FIG.
3 is a flowchart illustrating a movie recommendation method according to an exemplary embodiment of the present invention.
4 is a flowchart illustrating a process of generating a list of related movies according to an exemplary embodiment of the present invention.
5 is a view showing an example of first rating data of users for a movie.
6 is a diagram showing a fuzzy membership function for generating second rating data according to the present invention.
FIG. 7 is a diagram showing the first rating data shown in FIG. 5 using the fuzzy membership value and the second rating data using the fuzzy membership function of FIG. 6 (a).

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description and the accompanying drawings, detailed description of well-known functions or constructions that may obscure the subject matter of the present invention will be omitted. It should be noted that the same constituent elements are denoted by the same reference numerals as possible throughout the drawings.

The terms and words used in the present specification and claims should not be construed to be limited to ordinary or dictionary meanings and the inventor is not limited to the concept of terminology for describing his or her invention in the best way. It should be interpreted as meaning and concept consistent with the technical idea of the present invention. Therefore, the embodiments described in the present specification and the configurations shown in the drawings are merely the most preferred embodiments of the present invention, and not all of the technical ideas of the present invention are described. Therefore, It is to be understood that equivalents and modifications are possible. Also, the terms first, second, etc. are used for describing various components and are used only for the purpose of distinguishing one component from another component, and are not used to define the components.

1 is a block diagram showing a movie recommendation apparatus according to an embodiment of the present invention.

Referring to FIG. 1, a recommendation apparatus 100 according to an exemplary embodiment of the present invention may include a data acquisition unit 110, an association list generation unit 120, and a movie recommendation unit 130.

The data acquisition unit 110 may acquire first rating data including a user's rating for the movie.

In this case, the first rating data may include information that can be obtained from the input log, that is, a user ID that can identify a user, a movie ID that can identify a movie, and a rating in numerical form. Also, if the user ID, movie ID, and rating can be extracted from other types of input logs, first rating data can be obtained. For example, in order to process the first rating data as one transaction, it can be represented as 'User1 (m1, r_1_1), (m3, r_1_3), ..., (m100, r_1_100)'. Here, User1 can display the user ID, m1, m3, and m100 as movie IDs, r_1_1, r_1_3, and r_1_100 as ratings, and nth user can display ratings of mth movies as r_n_m.

Also, the association list generation unit 120 converts the acquired first rating data to second rating data, applies the converted second rating data to fuzzy association rule mining, and generates an associated movie list can do.

In this case, fuzzy association rule mining is a technique applied to association rule mining by fuzzy theory, and it can be expressed from the binary logic that each object belongs to a certain set or does not belong to each subject, . Therefore, it is possible to calculate the association between the products registered in the market or the store by calculating the association between the consumed products based on the user's log, such as association rule mining.

Such a fuzzy association rule is mainly calculated by using a single machine. For the recommendation of a large capacity, it is necessary to distribute the logic. In this case, the present invention uses fuzzy association rule mining more effectively You can use Hadoop's MapReduce to compute it.

In this case, in MapReduce, it is possible to solve by defining <key, value> for each step of mapper and reducer. <key, value> is a data pair that is the basic unit in which data is processed, and key and value can be defined as arbitrary structures or classes to process complex types of data.

At this time, a fuzzy membership value is acquired by assigning a rating to at least one of a fuzzy membership function including a triangular membership function, a trapezoidal membership function, and a Gaussian membership function , The language label according to the obtained fuzzy membership value can be converted into the second rating data by replacing with the rating.

At this time, if the rating included in the first rating data is substituted into the fuzzy membership function, the fuzzy membership value can be obtained with a value between 0 and 1 in relation to the language label, and the value of the language label having a large fuzzy membership value Can be converted into the second rating data by substituting the score with the rating. For example, a fuzzy membership value of 0.3 is obtained with respect to a language label having a value obtained by generating a fuzzy membership value using a fuzzy membership function of eight points included in the first rating data, If the fuzzy belonging value of 0.7 is obtained in association with the 'good' language label, the second rating data can be generated by replacing the information about the rating point 8 of the first rating data with 'good'.

In this case, each fuzzy membership function is associated with a language label such as 'No,' 'Normal', 'Good', and the range of the fuzzy membership function and the number of fuzzy membership functions can be specified separately.

The association list generation unit 120 generates at least one of fuzzy reliability and fuzzy correlation using fuzzy association rule mining and generates a related movie list based on at least one of the generated fuzzy reliability and fuzzy correlation .

In addition, the association list generation unit 120 may combine the converted second rating data according to the fuzzy association rule to generate a movie-specific association combination, generate a rating history for each movie organized by the second rating data for each movie , And the fuzzy support for each movie can be calculated using the generated rating history of each movie.

At this time, the film-by-film association combination can generate the combination with the length of the association rule preset according to the fuzzy association rule. For example, if the length of the association rule is 2, a movie combination for movies m1 and m3 is created with 'm1, m3, user1, r_1_1, r_1_3' data such as 'm1, m3, (user1, r_1_1, r_1_3), (uesr7, r_7_1, r_7_3), ..., (userN, r_N_1, r_N_3)' can be collected. In this case, the expression such as r_n_m can be interpreted as a rating value of the n-th user for the m-th movie.

The rating history of each movie is obtained by collecting second rating data of the form of 'User1 (m1, r_1_1), (m3, r_1_3), ..., (m100, r_1_100) (user1, r_1_1), (user1, r_2_1), ..., (userN, r_N_1) '.

는 l번째 소속 함수에 대해

값이 가지는 소속도 값,

는 정규화된 기본값 및

는 트랜잭션 DB

에서 i번째 레코드 값을 나타낼 수 있다.At this time, the fuzzy support per movie can be calculated using the reference value obtained by normalizing the fuzzy membership value. For example, each fuzzy membership value can be normalized using the following equation (1). At this time, in Equation (1)

For the lth membership function

The value belonging to the value,

Is the normalized default and

The transaction DB

Can represent the i-th record value.

In this way, the fuzzy support for each movie can be calculated using the reference value obtained by normalizing the fuzzy membership value as shown in the following equation.

는 l번째 소속 함수에 대해

값이 가지는 소속도 값,

는 정규화된 기본값,

는 트랜잭션 DB

에서 i번째 레코드 값 및

는 계산한 기준 값을 대입하여 계산한 퍼지 지지도를 나타낼 수 있다.Further, the fuzzy degree of support can be calculated using the following equation (2). At this time, in Equation (2)

For the lth membership function

The value belonging to the value,

Is the normalized default,

The transaction DB

The value of the ith record in

Can represent the fuzzy support calculated by substituting the calculated reference value.

Also, fuzzy reliability can be calculated by using at least two of the fuzzy membership functions to calculate the association fuzzy support for the movie association combination, and using at least one of the fuzzy support for the movie and the associated association fuzzy support.

는 l번째 소속 함수에 대해

값이 가지는 소속도 값,

는 정규화된 기본값,

는 트랜잭션 DB

에서 i번째 레코드 값,

,

는 기준 값을 대입하여 계산한 퍼지 지지도 및

는 퍼지 신뢰도를 나타낼 수 있다.In this case, the associative combination fuzzy support can be expressed as 'm1, m2, MF_1, MF_2, FS (m1, MF1, m2, MF_2)' where m is movie, MF is fuzzy membership function, Lt; / RTI &gt; For example, fuzzy reliability can be calculated using Equation 3 below. At this time, in Equation (3)

For the lth membership function

The value belonging to the value,

Is the normalized default,

The transaction DB

The i-th record value,

,

Is the fuzzy support calculated by substituting the reference value and

Can represent fuzzy reliability.

는 l번째 소속 함수에 대해

값이 가지는 소속도 값,

는 정규화된 기본값,

는 트랜잭션 DB

에서 i번째 레코드 값,

,

는 기준 값을 대입하여 계산한 퍼지 지지도,

는 퍼지 신뢰도 및

는 퍼지 상관도를 나타낼 수 있다.Also, the fuzzy correlation can be calculated using at least one of the squared values of the per-movie fuzzy support, fuzzy reliability, and per-movie fuzzy support. For example, the fuzzy correlation can be calculated using Equation (4) below. At this time, in Equation 4,

For the lth membership function

The value belonging to the value,

Is the normalized default,

The transaction DB

The i-th record value,

,

The fuzzy degree of support calculated by substituting the reference value,

Fuzzy reliability and

Can represent a fuzzy correlation.

here

The movie recommendation unit 130 may recommend the movie to the recommendation target user using the related movie list. For example, it is possible to recommend movies in the order of movies more suitable for the user by showing them to the recommendation target users in order of movies having high relevance among the movies included in the related movie list.

At this time, it is possible to determine the ranking of the associated movie list generated according to the predetermined importance, and to recommend the movie in the order of the determined associated high movie list. For example, using the language label included in the second rating data, the relationship between movies is good -> good or the relation of 'nope -> good' The relationship of 'good -> no,' 'no ->, and' no 'can be used for inducing curiosity rather than inducing direct purchase. In addition, the relation that is connected to 'normal' may be unnecessary information in terms of recommendation service because it is difficult to link with direct recommendation service function. Thus, the relationship between 'good -> good' and 'not - good' is determined to be of high importance, and the relation of 'good -> dislike', 'dislike -> dislike' It can be determined that the order of importance is low.

Also, if there are duplicate movies between the associated movie lists, the duplicate movies in the sublist can be deleted based on the ranking of the associated movie lists. For example, if there is a movie B in the list of movies that says "a movie user A is good," and there is a movie B in a movie list that says "he / she does not like a movie user A" It is possible to delete the movie B from the list of movies in which a user who does not like the movie A is said to be relatively inferior to the user by checking the ranking.

By using the movie recommendation device, users can recommend a movie suitable for the user's preference by recommending a movie related to the user through rating information about the movie.

FIG. 2 is a block diagram illustrating an association list generation unit of the movie recommendation apparatus of FIG. 1. FIG.

Referring to FIG. 2, the association list generation unit 120 of the movie recommendation apparatus of FIG. 1 includes an association combination generation unit 210 and a fuzziness degree calculation unit 220.

The association combination generation unit 210 may combine the converted second rating data according to the fuzzy association rule to generate a movie-specific association combination.

At this time, the film-by-film association combination can generate the combination with the length of the association rule preset according to the fuzzy association rule. For example, if the length of the association rule is 2, a movie combination for movies m1 and m3 is created with 'm1, m3, user1, r_1_1, r_1_3' data such as 'm1, m3, (user1, r_1_1, r_1_3), (uesr7, r_7_1, r_7_3), ..., (userN, r_N_1, r_N_3)' can be collected. In this case, the expression such as r_n_m can be interpreted as a rating value of the n-th user for the m-th movie.

The fuzzy confidence calculation unit 220 may generate a rating history for each movie, which is obtained by classifying the second rating data for each movie, and calculate the fuzzy support for each movie using the rating history for each movie.

In this case, the rating history of each movie is obtained by collecting second rating data of the form of 'User1 (m1, r_1_1), (m3, r_1_3), ..., (m100, r_1_100) , (user1, r_1_1), (user1, r_2_1), ..., (userN, r_N_1) '.

At this time, the fuzzy support per movie can be calculated using the reference value obtained by normalizing the fuzzy membership value. For example, each fuzzy belonging value can be normalized using Equation 1 described above.

In this way, the fuzzy support for each movie can be calculated using the reference value obtained by normalizing the fuzzy membership value as shown in the following equation.

In addition, the fuzzy degree of support can be calculated using Equation 2 described above.

3 is a flowchart illustrating a movie recommendation method according to an exemplary embodiment of the present invention.

Referring to FIG. 3, a movie recommendation method according to an embodiment of the present invention may acquire first rating data including a rating of a user for a movie (S310).

In this case, the first rating data may include information that can be obtained from the input log, that is, a user ID that can identify a user, a movie ID that can identify a movie, and a rating in numerical form. Also, if the user ID, movie ID, and rating can be extracted from other types of input logs, first rating data can be obtained. For example, in order to process the first rating data as one transaction, it can be represented as 'User1 (m1, r_1_1), (m3, r_1_3), ..., (m100, r_1_100)'. Here, User1 can display the user ID, m1, m3, and m100 as movie IDs, r_1_1, r_1_3, and r_1_100 as ratings, and nth user can display ratings of mth movies as r_n_m.

In addition, the movie recommendation apparatus according to an embodiment of the present invention converts the acquired first rating data to second rating data, applies the converted second rating data to a fuzzy association rule mining A movie list can be generated (S320).

In this case, fuzzy association rule mining is a technique applied to association rule mining by fuzzy theory, and it can be expressed from the binary logic that each object belongs to a certain set or does not belong to each subject, . Therefore, it is possible to calculate the association between the products registered in the market or the store by calculating the association between the consumed products based on the user's log, such as association rule mining.

Such a fuzzy association rule is mainly calculated by using a single machine. For the recommendation of a large capacity, it is necessary to distribute the logic. In this case, the present invention uses fuzzy association rule mining more effectively You can use Hadoop's MapReduce to compute it.

In this case, in MapReduce, it is possible to solve by defining <key, value> for each step of mapper and reducer. <key, value> is a data pair that is the basic unit in which data is processed, and key and value can be defined as arbitrary structures or classes to process complex types of data.

At this time, a fuzzy membership value is acquired by assigning a rating to at least one of a fuzzy membership function including a triangular membership function, a trapezoidal membership function, and a Gaussian membership function , The language label according to the obtained fuzzy membership value can be converted into the second rating data by replacing with the rating.

At this time, if the rating included in the first rating data is substituted into the fuzzy membership function, the fuzzy membership value can be obtained with a value between 0 and 1 in relation to the language label, and the value of the language label having a large fuzzy membership value Can be converted into the second rating data by substituting the score with the rating. For example, a fuzzy membership value of 0.3 is obtained with respect to a language label having a value obtained by generating a fuzzy membership value using a fuzzy membership function of eight points included in the first rating data, If the fuzzy belonging value of 0.7 is obtained in association with the 'good' language label, the second rating data can be generated by replacing the information about the rating point 8 of the first rating data with 'good'.

In this case, each fuzzy membership function is associated with a language label such as 'No,' 'Normal', 'Good', and the range of the fuzzy membership function and the number of fuzzy membership functions can be specified separately.

The association list generation unit 120 generates at least one of fuzzy reliability and fuzzy correlation using fuzzy association rule mining and generates a related movie list based on at least one of the generated fuzzy reliability and fuzzy correlation .

In addition, the association list generation unit 120 may combine the converted second rating data according to the fuzzy association rule to generate a movie-specific association combination, generate a rating history for each movie organized by the second rating data for each movie , And the fuzzy support for each movie can be calculated using the generated rating history of each movie.

At this time, the film-by-film association combination can generate the combination with the length of the association rule preset according to the fuzzy association rule. For example, if the length of the association rule is 2, a movie combination for movies m1 and m3 is created with 'm1, m3, user1, r_1_1, r_1_3' data such as 'm1, m3, (user1, r_1_1, r_1_3), (uesr7, r_7_1, r_7_3), ..., (userN, r_N_1, r_N_3)' can be collected. In this case, the expression such as r_n_m can be interpreted as a rating value of the n-th user for the m-th movie.

The rating history of each movie is obtained by collecting second rating data of the form of 'User1 (m1, r_1_1), (m3, r_1_3), ..., (m100, r_1_100) (user1, r_1_1), (user1, r_2_1), ..., (userN, r_N_1) '.

At this time, the fuzzy support per movie can be calculated using the reference value obtained by normalizing the fuzzy membership value. For example, each fuzzy belonging value can be normalized using Equation 1 described above.

In this way, the fuzzy support for each movie can be calculated using the reference value obtained by normalizing the fuzzy membership value as shown in the following equation.

In addition, the fuzzy degree of support can be calculated using Equation 2 described above.

Also, fuzzy reliability can be calculated by using at least two of the fuzzy membership functions to calculate the association fuzzy support for the movie association combination, and using at least one of the fuzzy support for the movie and the associated association fuzzy support.

In this case, the associative combination fuzzy support can be expressed as 'm1, m2, MF_1, MF_2, FS (m1, MF1, m2, MF_2)' where m is movie, MF is fuzzy membership function, Lt; / RTI &gt; For example, the fuzzy reliability can be calculated using Equation 3 described above.

Also, the fuzzy correlation can be calculated using at least one of the squared values of the per-movie fuzzy support, fuzzy reliability, and per-movie fuzzy support. For example, the fuzzy correlation can be calculated using Equation 4 described above.

In addition, the movie recommendation method according to an embodiment of the present invention may recommend a movie using a related movie list to a recommendation target user (S330). For example, it is possible to recommend movies in the order of movies more suitable for the user by showing them to the recommendation target users in order of movies having high relevance among the movies included in the related movie list.

At this time, it is possible to determine the ranking of the associated movie list generated according to the predetermined importance, and to recommend the movie in the order of the determined associated high movie list. For example, using the language label included in the second rating data, the relation between movies is good -> good or the relation of 'nope -> good' The relationship between 'good -> nope' and 'nope -> nope' can be used for inducing curiosity rather than inducing direct purchase. In addition, the relationship that is connected to 'normal' may be unnecessary information in terms of service because it is difficult to connect with the function of direct recommendation service. Therefore, the relationship between 'good -> good' and 'bad - good' is determined to be of high importance, and the relationship of 'good -> no,' 'no ->, It can be determined as the order of importance.

Also, if there are duplicate movies between the associated movie lists, the duplicate movies in the sublist can be deleted based on the ranking of the associated movie lists. For example, if there is a movie B in the list of movies that says "a movie user A is good," and there is a movie B in a movie list that says "he / she does not like a movie user A" It is possible to delete the movie B from the list of movies in which a user who does not like the movie A is said to be relatively inferior to the user by checking the ranking.

Through such a movie recommendation method, it is possible to provide a reliable movie recommendation service using a large-capacity user log to users who use the movie recommendation service.

4 is a flowchart illustrating a process of generating a list of related movies according to an exemplary embodiment of the present invention.

Referring to FIG. 4, the process of generating a list of related movies according to an exemplary embodiment of the present invention may acquire second rating data by replacing a rating included in the first rating data with a language label (S410).

At this time, a fuzzy membership value is acquired by assigning a rating to at least one of a fuzzy membership function including a triangular membership function, a trapezoidal membership function, and a Gaussian membership function , The language label according to the obtained fuzzy membership value can be converted into the second rating data by replacing with the rating.

In addition, in the process of generating the associated movie list according to an embodiment of the present invention, the converted second rating data may be combined according to the fuzzy association rule to generate a movie-specific association combination (S420).

In addition, the process of generating a list of associated movies according to an embodiment of the present invention generates a rating history for each movie in which the second rating data is sorted for each movie (S430) The support can be calculated (S440).

In addition, the process of generating the associated movie list according to an embodiment of the present invention may be performed by using the association combinations and the per-movie fuzzy support generated and calculated in steps S420 and S440, At least two or more combinations may be combined to calculate the association fuzzy support for the movie-specific association (S450).

In addition, the process of generating the associated movie list according to an exemplary embodiment of the present invention may calculate the fuzzy reliability using at least one of the fuzzy support of the movie and the calculated association fuzzy support (S460).

In addition, the process of generating the associated movie list according to an exemplary embodiment of the present invention may calculate the fuzzy correlation using one or more of the fuzzy support, fuzzy reliability, and per-movie fuzzy support scores at step S470.

In addition, the process of generating an associated movie list according to an exemplary embodiment of the present invention may generate an associated movie list based on one of the generated fuzzy reliability and fuzzy correlation (S480).

5 is a view showing an example of first rating data of users for a movie.

6 is a diagram showing a fuzzy membership function for generating second rating data according to the present invention.

FIG. 7 is a diagram showing the first rating data shown in FIG. 5 using the fuzzy membership value and the second rating data using the fuzzy membership function of FIG. 6 (a).

5, 6, and 7, when a user gives a rating to a movie as shown in FIG. 5, the fuzzy membership value shown in FIG. 6 and the second rating data Lt; / RTI &gt;

For example, in the case of using the fuzzy membership function in FIG. 6A, it can be seen that the user 1 in FIG. 5 assigns two points to the movie 1. At this time, the value of the fuzzy belonging to the movie 1 of the user 1 is found to be unpleasant: 1.0, normal: 0.0, and good: 0.0 in FIG. Accordingly, in this case, it is possible to replace the language label 'No' with the second rating data of the first rating data to convert it into second rating data.

Also, it can be seen that the user 1 has given 8 points to the movie 4. At this time, the value of the fuzzy belonging speed for the movie 4 of the user 1 is shown as "No, 0.0, Normal: 0.33, and Good: 0.67" in FIG. Therefore, in this case, it is possible to convert the language label 'good' into the second rating data by replacing 8 points of the first rating data.

The movie recommendation method according to the present invention may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions recorded on the medium may be those specially designed and constructed for the present invention or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks, and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Includes all types of hardware devices that are specially configured to store and execute magneto-optical media and program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions may include machine language code such as those generated by a compiler, as well as high-level language code that may be executed by a computer using an interpreter or the like. Such a hardware device may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

As described above, the apparatus and method for recommending a movie based on the distributed fuzzy association rules mining according to the present invention can be applied to the configurations and methods of the embodiments described above in a limited manner, All or some of the embodiments may be selectively combined.

According to the present invention, the first rating data including the rating for the movie is converted into the second rating data, the second rating data is applied to the fuzzy association rule mining to generate the associated movie list, It is possible to effectively recommend a movie suitable for the recommendation target user's recommendation. Furthermore, since the data processing method suitable for a distributed framework is used to apply such a recommendation function to a large capacity user, it can be applied to a large scale log data, thereby providing a more reliable service.

100: Movie recommendation device 110:
120: Association list generation unit 130: Movie recommendation unit
210: associative combination generation unit 220:

Claims (17)

A data acquiring unit acquiring first rating data including a rating for a movie;
An association list generation unit for converting the acquired first rating data to second rating data and applying the converted second rating data to a fuzzy association rule mining to generate an associated movie list; And
A movie recommendation section for recommending a movie using the associated movie list to a recommendation target user,
Wherein the motion recommendation apparatus comprises: The method according to claim 1,
The association list generation unit
A fuzzy membership function including a triangular membership function, a trapezoidal membership function, and a Gaussian membership function to obtain a fuzzy membership value by substituting the rating into one or more fuzzy membership functions,
And converting the language label according to the acquired fuzzy membership value into the second rating data by replacing the language label with the rating. The method according to claim 1,
The association list generation unit
Generating at least one of fuzzy reliability and fuzzy correlation using the fuzzy association rules mining,
And generates the associated movie list based on at least one of the generated fuzzy reliability and the fuzzy correlation. The method of claim 3,
The associated movie list generation unit
An associative combination generating unit for generating a movie-specific association by combining the converted second rating data according to a fuzzy association rule; And
And a fuzzy support calculation unit for generating a fuzzy history for each movie in which the second rating data is organized for each movie and calculating fuzzy support for each movie using the generated rating history for each movie. The method of claim 4,
The purging support calculation unit
Wherein the fuzzy degree of support for each movie is calculated using a reference value obtained by normalizing the value of the fuzzy support value. The method of claim 4,
The association list generation unit
Wherein the fuzzy membership function is a combination of at least two of the fuzzy membership functions,
Wherein the fuzzy reliability calculation unit calculates the fuzzy reliability using at least one of the per-movie fuzzy support and the calculated associative combination fuzzy support. The method of claim 4,
The association list generation unit
Wherein the fuzzy correlation calculating unit calculates the fuzzy correlation using at least one of a squared value of the per-movie fuzzy support, fuzzy confidence, and per-movie fuzzy support. The method according to claim 1,
The movie recommendation unit
Determining a ranking of the associated movie list generated according to a predetermined importance, and recommending the movie in the order of the determined associated high movie list. The method according to claim 1,
The associated movie list
A movie title, a genre, a director, a country, a year of production, and an image. Obtaining first rating data including a rating for a movie;
Converting the acquired first rating data to second rating data, applying the converted second rating data to fuzzy association rule mining to generate a list of associated movies; And
Recommending a movie to the recommendation target user using the associated movie list
Wherein the movie recommendation method comprises: The method of claim 10,
The step of generating the associated movie list
Obtaining a fuzzy membership value by substituting the rating into at least one of a fuzzy membership function including a triangular membership function, a trapezoidal membership function, and a Gaussian membership function, Including,
And converting the language label according to the acquired fuzzy membership value into the second rating data by replacing the rating label with the rating value. The method of claim 10,
The step of generating the associated movie list
Generating at least one of fuzzy reliability and fuzzy correlation using the fuzzy association rule mining; And
And generating the associated movie list based on at least one of the generated fuzzy reliability and the fuzzy correlation. The method of claim 12,
The step of generating the associated movie list
Combining the converted second rating data according to the fuzzy association rule to generate a movie-specific association combination; And
Generating a rating score for each movie in which the second rating data is organized for each movie, and calculating fuzzy support for each movie using the rating history for each movie. 14. The method of claim 13,
The step of generating the associated movie list
Calculating at least two of the fuzzy membership functions to associate the combined fuzzy support for the per-movie association combination; And
And calculating the fuzzy reliability using at least one of the per-movie fuzzy support and the calculated associative combination fuzzy support. 14. The method of claim 13,
The step of generating the associated movie list
And calculating the fuzzy correlation using at least one of squared values of fuzzy support, fuzzy reliability, and per-movie fuzzy support of the movie. The method of claim 10,
The step of recommending the movie
Determining a ranking of the generated associated movie list according to a predetermined importance,
And recommending the movie in the order of the determined associated high movie list. A computer-readable recording medium on which a program for executing the method according to any one of claims 10 to 16 is recorded.

Images