KR102403681B1 - System and Method for Inducing Intelligent Fuzzy Based on User Loacation and Keyword - Google Patents

Abstract

The user location and keyword-based intelligent fuzzy induction system and method have the effect of using a location-based social network to automatically recommend cultural sites that are highly relevant to the user's preferences and interests to inform interesting and new cultural sites.

Description

System and Method for Inducing Intelligent Fuzzy Based on User Location and Keyword}

The present invention relates to an intelligent fuzzy induction system and method, and more particularly, to a user location and keyword-based intelligent fuzzy induction system for automatically recommending cultural sites highly relevant to a user's preferences and interests using a location-based social network; it's about how

The development of communication technologies and web-based services can generate large amounts of information.

A recent Recommender System uses a cloud computing paradigm to discover interesting and new cultural sites based on user preferences and interests.

Since most users tend to visit only existing cultural sites, there is a problem in that attractive cultural sites cannot be found and hidden due to lack of awareness of cultural sites.

The recommendation algorithm is effectively used in various fields such as movies, advertisements, e-commerce, music, e-books, and travel, thereby proving its effectiveness.

However, in the field of cultural relics, the recommendation algorithm has not been used more effectively.

Korea Registered Patent No. 10-1382069

In order to solve such a problem, the present invention provides a user location and keyword-based intelligent fuzzy induction system and method for automatically recommending a cultural site highly relevant to a user's preference and interest using a location-based social network. There is this.

User location and keyword-based intelligent fuzzy induction system according to the features of the present invention for achieving the above object,

a data set collection unit that collects information related to cultural sites by using a social network data set;

Receives the cultural relic-related information from the data set collection unit, extracts candidate keywords related to cultural relics using RAKE (Rapid Automatic Keyword Extraction), a keyword extraction algorithm, and ranks according to the frequency of each extracted candidate keyword a keyword feature extraction unit for selecting preset top N candidate keywords as a keyword list;

a classification unit for grouping and clustering target users in a certain range using a Fuzzy-KNN (K-Nearest Neighbor) model that performs learning based on similarity in the selected keyword list; and

and a cultural site recommendation module for generating recommendation information by ranking cultural sites using user location and user preference for social activities in the clustered keyword list.

A user location and keyword-based intelligent fuzzy induction method according to a feature of the present invention,

collecting cultural site-related information using the social network data set;

Candidate keywords related to cultural relics are extracted from the collected cultural relic-related information using RAKE (Rapid Automatic Keyword Extraction), a keyword extraction algorithm, and ranked according to the frequency of each extracted candidate keyword. selecting a candidate keyword as a keyword list;

clustering and grouping target users in a certain range using a Fuzzy-KNN (K-Nearest Neighbor) model that performs learning based on similarity in the selected keyword list; and

Ranking the collected cultural relics using user location and user preference for social activity in the clustered keyword list, and generating customized recommendation information for preset N high-ranking cultural relics according to user preferences and interests It is characterized in that it comprises a step.

According to the above-described configuration, the present invention has an effect of automatically recommending cultural relics highly relevant to the user's preference and interest using a location-based social network to inform interesting and new cultural relics.

1 is a diagram showing the configuration of an intelligent fuzzy induction system based on user location and keywords according to an embodiment of the present invention.
2 is a block diagram schematically illustrating an internal configuration of a recommendation server according to an embodiment of the present invention.
3 is a diagram illustrating a process of generating customized recommendation information for recommending cultural relics according to an embodiment of the present invention.
4 is a diagram showing precision, recall, and f-measurement for four data sets (Foursquare, Gowalla, Yelp, TripAdvior) according to an embodiment of the present invention.
5 is a diagram illustrating recommendation accuracy of a recommendation server evaluated in four data sets (Foursquare, Gowalla, Yelp, TripAdvior) according to an embodiment of the present invention.

Throughout the specification, when a part "includes" a certain element, it means that other elements may be further included, rather than excluding other elements, unless otherwise stated.

The present invention proposes a personalized recommendation model in the field of cultural heritage with the help of a cloud computing environment.

The recommendation system of the present invention may show a technology for recommending personalized cultural heritage in the field of cultural heritage tourism service.

1 is a diagram showing the configuration of an intelligent fuzzy induction system based on user location and keywords according to an embodiment of the present invention, FIG. 2 is a block diagram schematically showing the internal configuration of a recommendation server according to an embodiment of the present invention, FIG. 3 is a diagram illustrating a process of generating customized recommendation information for recommending cultural relics according to an embodiment of the present invention.

The intelligent fuzzy induction system 100 based on user location and keyword according to an embodiment of the present invention includes a user terminal 110 , a communication network 111 , and a recommendation server 120 .

When the user terminal 110 approaches the cultural relic of a specific location, it generates location information and transmits it to the recommendation server 120 through the communication network 111 .

When receiving the location information from the user terminal 110 , the recommendation server 120 generates recommendation information including a list of cultural relics based on the location information and transmits it to the user terminal 110 through the communication network 111 .

The recommendation server 120 periodically receives location information from the user terminal 110, and when the location of the user terminal 110 is changed, the customized recommendation information is automatically updated to include new cultural sites in the recommendation information. .

The recommendation server 120 according to an embodiment of the present invention includes a data set collection unit 121, a keyword feature extraction unit 122, a Fuzzy-KNN classification unit 123, a cultural site recommendation module 124, and a control unit 125. and a wireless communication unit 126 .

The data set collecting unit 121 collects information related to cultural relics by using social network data sets such as Foursquare, Gowalla, Yelp, and TripAdvior.

This social network data set is collected in depth to extract the top N keyword sets.

Foursquare, Gowalla, Yelp, and TripAdvior are some of the popular location-based social networks.

A social network may contain information such as a user's social ties, as well as daily activities, dependencies, and social affinity between users.

The cultural site recommendation system 100 of the present invention uses a social network based on the current location.

The cultural site recommendation system 100 integrates social activity with geographic location and user preferences to generate recommendation information for cultural sites.

The cultural site recommendation system 100 provides a cultural site recommendation service using cloud computing. Cloud computing facilitates powerful and flexible storage and computing services by virtualizing and converging information from disparate sources.

Cloud computing is the interaction between entities such as service providers, consumers and brokers to provide better services. In the field of cultural sites, cloud computing can be used to reveal the unique value of some unknown cultural sites to the public.

The data set collection unit 121 collects useful information of the target user from the social network profile. Social network data sets include information such as check-in times, social connections, reviews, comment ratings, and feedback. Here, the target user may be the user terminal 110 .

The social network data set also includes information on places that have been previously visited to help predict ratings for unvisited cultural sites.

80% of the data set is considered training data and the remaining 20%s are considered test data.

The keyword feature extraction unit 122 collects all accessible user profiles (information related to cultural relics), and then extracts candidate keywords related to cultural relics using Rapid Automatic Keyword Extraction (RAKE), which is a language-independent keyword extraction algorithm.

Keyword-based user clustering is implemented using a Fuzzy-KNN (K-Nearest Neighbor) clustering model.

The keyword feature extraction unit 122 uses the RAKE method in the user profile to generate a set of candidate keywords related to cultural relics from the extracted candidate keywords. Here, the user profile consists of information available in the user's social network, such as personal information, social relationships, location information, and user basic description.

The keyword feature extraction unit 122 extracts user preference information including posts, feedback, reviews, and evaluations of the user in order to generate a set of candidate keywords for cultural relics.

When all candidate keyword sets are generated, the keyword feature extraction unit 122 calculates the word score of each candidate keyword by calculating the frequency and degree of each candidate keyword as shown in Equation 1 below.

Here, Degree(CK) is the total number of times a candidate keyword appears in a post, review, or feedback, and Frequency(CK) indicates regularity of each candidate keyword.

The keyword feature extraction unit 122 selects preset top N candidate keywords as a keyword list.

An example of the keyword list extracted from the online review generated by the RAKE method is shown in Table 1 below.

The extracted keyword list is used to cluster target users into equal categories.

The keyword list is used as a function to cluster target users into a reliable range using the Fuzzy-KNN clustering method.

The Fuzzy-KNN classifier 123 performs learning based on similarity and does not generate a classification model until the test data is classified, so it is an instance-based learner.

Fuzzy-KNN algorithm selects the k nearest neighbor sets by calculating the distance between the test data and the training data set. Fuzzy-KNN algorithm improves accuracy by weighting or applying the concept of fuzzy sets.

The Fuzzy-KNN classifier 123 solves the classification problem accurately by integrating the fuzzy set theory and k nearest neighbors.

The Fuzzy-KNN classification unit 123 uses the corresponding class membership level for class assignment of test data. The class membership level is calculated by the initial class membership level and the final class membership level.

The initial class membership degree is determined using the class distribution of the k neighbor sets and the class information of the training data.

The final class membership degree is determined by the distance between the test data and the training data belonging to k neighboring sets and the initial class membership degree, and indicates how much the test data can belong to the existing class.

The fuzzy-KNN classifier 123 determines the k neighbor sets using the similarity between the training data and the test data.

In the similarity measurement, after calculating the similarity between the target user and other existing users, the target users are clustered based on the extracted keyword list using Fuzzy-KNN.

The Fuzzy-KNN classifier 123 assigns a class membership function of the data point distance from k nearest neighbors. Each data point gets a unique value for each neighbor.

More closely related neighbors are grouped into a single cluster, and a single data point may belong to more than one cluster.

When the target user shows a high degree of class membership, the intensity of the user belonging to the specific cluster is also high.

The membership function of the data point of the Fuzzy-KNN classifier 123 is estimated as in Equation 2 below.

는 i번째 클래스에 대한 사용자 x의 퍼지 클래스 멤버쉽이고,

는 k 최근접 이웃들 중에 사용자 x의 키워드 j의 클래스 멤버쉽 정도이다.here,

is the fuzzy class membership of user x to the i-th class,

is the degree of class membership of the keyword j of user x among k nearest neighbors.

는 클래스 i에 대한 x의 멤버쉽 정도를 의미한다.

is the degree of membership of x in class i.

Variables i = 1 to c represent the number of classes, j = 1 to k are the predefined number of neighbors, k is the number of class sets, and m is a constant parameter, always between 1 and infinity.

로 추정된다.and the Euclidean distance between the target user x and the j-th nearest neighbor

is estimated to be

The Fuzzy-KNN classifier 123 clusters the target users in a reliable range by using the Fuzzy-KNN clustering method from the extracted keyword list.

The cultural site recommendation system 100 may generate personalized cultural site recommendation information through a rating matrix.

Once users are grouped into appropriate clusters, a rating matrix is constructed from ratings provided by other neighbors of the cluster for various Cultural Heritage (CH) resources.

The cultural site recommendation module 124 predicts a rating for a cultural site not visited by the target user based on the rating behavior of the nearest neighbor that is more similar to the rating activity of the target user.

Table 2 below shows an example of a rating matrix.

In general, similarity measurement and user clustering are two widely used approaches for finding neighboring neighbors for class prediction.

The aforementioned Fuzzy-KNN classifier 123 utilizes the Fuzzy-KNN clustering model because the clustering model generates better results than the similarity-based model. In the present invention, if the Fuzzy-KNN clustering model is used, computational complexity can be greatly reduced by selecting k nearest neighbors instead of selecting all users in the cluster.

The cultural site recommendation module 124 calculates using the Maximizing Average Satisfaction method as in Equation 3 below to predict the grade of the cultural site that has not been visited.

The maximum average satisfaction method computes the average rating of all neighborhoods for a specific cultural site.

The cultural relics recommendation module 124 ranks the predicted cultural relics according to the predicted grade.

Assuming that the pre-set N high-ranking superlative cultural heritage is more related to the preference of the target user, the superlative cultural heritage may be suggested.

The maximum average satisfaction level used for predicting the level of cultural sites is defined by Equation 3 below.

는 i번째 문화 유적지, j=1 내지 k는 k 최근접 이웃들이고, Rating은 모든 k 최근접 이웃들의 i번째 문화 유적지를 위한 등급이다.where u is the target user,

is the i-th cultural site, j=1 through k are k nearest neighbors, and Rating is the rating for the i-th cultural site of all k nearest neighbors.

When the cultural site recommendation module 124 groups the target user into an appropriate cluster, the rating for the cultural site that has not been visited is predicted using k nearest neighbors and the Maximizing Average Satisfaction (MAS) method.

The cultural relics recommendation module 124 ranks the predicted cultural relics in order, and generates customized recommendation information according to user preferences and interests for the preset N high-grade cultural relics.

The recommendation server 120 may generate recommendation information by mining the rating and “like” of the active user in the social network item.

The recommendation server 120 may infer the current location, information such as explicit posts and reviews, interests and preferences related to cultural sites, by using the target user's social network data.

The recommendation server 120 may evaluate the rating and opinion of the target user on the cultural site that has already been visited for rating prediction.

The recommendation server 120 may rank the generated cultural relics in order according to the requirements and interests of the target user.

The recommendation server 120 generates personalized and customized recommendation information using social network data in a cloud computing environment.

When receiving the location information from the user terminal through the wireless communication unit 126, the control unit 125 generates recommendation information including a list of N high-grade cultural relics preset based on the location information to connect the wireless communication unit 126. through the user terminal 110 .

When the location of the user terminal 110 is changed, the control unit 125 automatically updates the recommended information to include new cultural sites in the recommended information, or transmits it to the user terminal 110 through the wireless communication unit 126 .

The control unit 125 uses a standard evaluation index to verify the effectiveness of the recommendation server 120 for recommending cultural relics.

Standard evaluation indicators are Recall, Precision, f-measurement, and Accuracy.

The precision is the ratio of positively predicted values, as shown in Equation 4 below, and is defined as the ratio of recommendations of cultural sites related to the user's interest.

rec _CH is the number of cultural sites proposed to the target user, and rel _CH is the number of actual cultural sites targeted to the target user's interest.

Recall (Recall) is defined as the ratio of appropriate cultural sites actually recommended, as in Equation 5 below.

The f-measurement is a harmonic average of precision, sensitivity, or recall, and is calculated by the following equation (6).

Accuracy is a decision evaluation metric that helps to generate a matching cultural site for a target user from a set of all available cultural sites based on their own requirements and interests, as shown in Equation 7 below.

Accuracy is an accurately predicted value, defined as the ratio between the positive (True Positive, TP) and negative (TN) and total cultural sites.

In order to improve the performance of the cultural site recommendation model, the evaluation prediction for unvisited cultural sites should be accurate and more relevant to users' preferences and interests.

The recommendation server 120 provides the target user's social information for the visited cultural sites, and predicts the top N cultural sites that have not been visited.

The user (or the user terminal 110 ) may experience the cultural relics generated by customized recommendations according to tastes and interests.

The user terminal 110 may receive distance information from the recommendation server 120 to the cultural site, images of other users, reviews, comments, and explanations and display them through a display unit (not shown).

The performance efficiency of the recommendation server 120 is evaluated by four data sets (Foursquare, Gowalla, Yelp, TripAdvior) of Recall, Precision, f-measure, and Accuracy. High precision means that the recommended cultural sites are highly related to users' interests and preferences, and high recall values indicate that they are not recommended as more relevant cultural sites.

As shown in Figure 4, precision, recall, and f-measures are presented for four data sets (Foursquare, Gowalla, Yelp, TripAdvior).

Recommendation accuracy is defined as the rate at which the recommended cultural sites are actually relevant to the user's interest. The higher the recommendation accuracy, the more likely the user will accept the recommendation.

As shown in FIG. 5 , the recommendation accuracy of the recommendation server 120 evaluated in four data sets (Foursquare, Gowalla, Yelp, TripAdvior) is shown.

In the above, the embodiment of the present invention is not implemented only through an apparatus and/or method, and may be implemented through a program for realizing a function corresponding to the configuration of the embodiment of the present invention, a recording medium in which the program is recorded, etc. And, such an implementation can be easily implemented by an expert in the technical field to which the present invention belongs from the description of the above-described embodiment.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements by those skilled in the art using the basic concept of the present invention as defined in the following claims are also provided. is within the scope of the right.

100: cultural site recommendation system 110: user terminal
111: communication network 120: recommendation server
121: data set collection unit 122: keyword feature extraction unit
123: Fuzzy-KNN Classification 124: Cultural Site Recommendation Module
125: control unit 126: wireless communication unit

Claims (10)

a data set collection unit that collects information related to cultural relics by using a social network data set;
Receives the cultural relic-related information from the data set collection unit, extracts candidate keywords related to cultural relics using RAKE (Rapid Automatic Keyword Extraction), a keyword extraction algorithm, and ranks according to the frequency of each extracted candidate keyword a keyword feature extraction unit for selecting preset top N candidate keywords as a keyword list;
a classification unit for grouping and clustering target users in a predetermined range using a Fuzzy-KNN (K-Nearest Neighbor) model that performs learning based on similarity in the selected keyword list; and
and a cultural site recommendation module for generating recommendation information by ranking cultural sites by integrating social activities with user location and user preference in the clustered keyword list. According to claim 1,
The candidate keyword set is a user location and keyword-based intelligent fuzzy induction system, characterized in that it includes user preference information including the user's posts, feedback, reviews, and ratings. According to claim 1,
When the candidate keyword set is generated, the keyword feature extraction unit calculates the word score of each candidate keyword by calculating the frequency and degree of each candidate keyword according to Equation 1 below, and each An intelligent fuzzy guidance system based on user location and keywords, characterized in that it ranks candidate keywords.
[Equation 1]

Here, Degree(CK) is the total number of times a candidate keyword appears in a post, review, or feedback, and Frequency(CK) indicates the regularity of each candidate keyword. According to claim 1,
The user location and keyword-based intelligent fuzzy induction system, characterized in that the cultural site recommendation module calculates the rating prediction of the cultural site by using Equation 2 below for a Maximizing Average Satisfaction method.
[Equation 2]

where u is the target user,

is the i-th cultural site, j=1 through k are k nearest neighbors, and Rating is the rating for the i-th cultural site of all k nearest neighbors. 5. The method of claim 4,
The cultural relics recommendation module ranks the predicted cultural relics in order, and generates customized recommendation information according to user preferences and interests to preset N high-grade cultural relics. and keyword-based intelligent fuzzy induction system. According to claim 1,
The cultural site recommendation module is a user location and keyword-based intelligent fuzzy induction system, characterized in that the rating for the cultural site not visited by the target user is predicted based on the rating behavior of the nearest neighbor that is more similar than the rating activity of the target user . According to claim 1,
When receiving the location information from the user terminal, based on the location information, further comprising a control unit for generating recommendation information including a list of N pre-set higher-grade cultural relics and transmitting it to the user terminal through a communication network,
The control unit is a user location and keyword-based intelligent fuzzy induction system, characterized in that when the location of the user terminal is changed, the recommendation information is automatically updated to include a new cultural relic in the recommendation information. The data set collecting unit may include: collecting information related to cultural relics by using a social network data set;
The keyword feature extraction unit extracts candidate keywords related to cultural heritage sites using RAKE (Rapid Automatic Keyword Extraction), which is a keyword extraction algorithm, from the collected cultural heritage site-related information, and ranks the extracted candidate keywords according to the frequency of each candidate keyword. selecting the set top N candidate keywords as a keyword list;
The fuzzy-KNN classifier comprises: clustering and grouping target users in a certain range using a Fuzzy-KNN (K-Nearest Neighbor) model that performs learning based on similarity in the selected keyword list; and
The cultural relics recommendation module ranks the collected cultural relics by integrating social activity with user location and user preferences in the clustered keyword list, and is customized according to user preferences and interests to preset N top-rated cultural relics. A user location and keyword-based intelligent fuzzy induction method comprising the step of generating recommended information. 9. The method of claim 8,
The step of selecting the keyword list is,
When the candidate keyword set is generated, the keyword feature extraction unit calculates the word score of each candidate keyword by calculating the frequency and degree of each candidate keyword according to Equation 1 below, and each User location and keyword-based intelligent fuzzy induction method, characterized in that it further comprises the step of ranking candidate keywords.
[Equation 1]

Here, Degree(CK) is the total number of times a candidate keyword appears in a post, review, or feedback, and Frequency(CK) indicates the regularity of each candidate keyword. 9. The method of claim 8,
The step of generating the customized recommendation information includes:
User location and keyword-based intelligent fuzzy induction, characterized in that the cultural site recommendation module further comprises calculating the rating prediction of the cultural site using Equation 2 below for a Maximizing Average Satisfaction method Way.
[Equation 2]

where u is the target user,

is the i-th cultural site, j=1 through k are k nearest neighbors, and Rating is the rating for the i-th cultural site of all k nearest neighbors.

Images