• The KIPS Transactions:PartB
• /
• v.14B no.7
• /
• pp.549-558
• /
• 2007

Collaborative filtering systems have problems when users rate items and the rated results depend on their feelings, as there is a possibility that the results include noise. The method proposed in this paper optimizes the matrix by excluding irrelevant ratings as information for recommendations from a user-item matrix using dispersion. It reduces the noise that results from predicting preferences based on original user ratings by inflecting the information for items and users on the matrix. The method excludes the ratings values of the utmost limits using a percentile to supply the defects of coefficient of variance and composes a weighted user-item matrix by combining the user coefficient of variance with the median of ratings for items. Finally, the preferences of the active user are predicted based on the weighted matrix. A large database of user ratings for movies from the MovieLens recommender system is used, and the performance is evaluated. The proposed method is shown to outperform earlier methods significantly.

• Journal of the Korean Data and Information Science Society
• /
• v.18 no.3
• /
• pp.659-668
• /
• 2007

Collaborative filtering approach for recommender systems are now widely applied in e-commerce to assist customers to find their needs from many that are frequently available. this approach makes recommendations for users based on the opinions to similar users in the system. But this approach is opened to users who present their preference to items or acquire the preference information form other users, noise in the system makes significant problem for accurate recommendation. In this paper, we analyze the relationship between the standard deviation of preference ratings for each user and the estimated ratings of them. The result shows that the possibility of the pre-filtering condition which detecting the factor of bad effect on the prediction of user's preference. It is expected that using this result will reduce the possibility of bad effect on recommender systems.

• Journal of KIISE:Databases
• /
• v.34 no.5
• /
• pp.389-395
• /
• 2007

Collaborative filtering is a popular technique for information filtering to reduce information overload and widely used in application such as recommender system in the E-commerce domain. Collaborative filtering systems collect human ratings and provide Predictions based on the ratings of other people who share the same tastes. The quality of predictions depends on the number of items which are commonly rated by people. Therefore, it is difficult to apply pure collaborative filtering algorithm directly to dynamic collections where items are constantly added or removed. In this paper we suggest a method for managing dynamic collections. It creates taste space for items using a technique called Singular Vector Decomposition (SVD) and maintains clusters of core items on the space to estimate relevance of past and future items. To evaluate the proposed method, we divide database of user ratings into those of old and new items and analyze predicted ratings of the latter. And we experimentally show our method is efficiently applied to dynamic collections.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.05a
• /
• pp.346-348
• /
• 2022

Media Contents Recommendation Web Service (service name 'mobodra') is a web service that analyzes media types and genre tastes for each user and recommends content accordingly. Users select some of the works randomly provided on the web when signing up for membership and analyze their tastes based on this. Based on this analysis, preferred content for each user is recommended. In this paper, we implement a content recommendation algorithm through item-based collaborative filtering. When the user's activity data or preference is re-examined, the above process is executed again to update the user's taste.

• Journal of Digital Convergence
• /
• v.14 no.8
• /
• pp.313-318
• /
• 2016

The recommendation service is changing from client-server based internet service to social networking. Especially in recent years, it is serving recommendations with personalization to users through crowdsourcing and social networking. The social networking based systems can be classified depending on methods of providing recommendation services and purposes by using memory and model based collaborative filtering. In this study, we proposed the social network based sensibility design recommendation using associative user. The proposed method makes {user - associative design} matrix through the social network and recommends sensibility design using the memory based collaborative filtering. For the performance evaluation of the proposed method, recall and precision verification are conducted. F-measure based on recommendation of social networking is used for the verification of accuracy.

• Journal of KIISE:Software and Applications
• /
• v.31 no.2
• /
• pp.153-163
• /
• 2004

Many recommender systems are based on Content-based Filtering and Social Filtering Both methods have their own advantages and disadvantages, and they complement each other rather than compete. So incorporating of both methods can make the better system and combination technique controls the quality of the entire recommender system. In this paper, we presented each user has his own tendency to decide which is the better recommendation for himself among the various recommendation results, and suggested the Personalized combination technique. To represent user tendency, we defined and used loyalty, diversity and pioneerity and showed by experiments that our combination technique is useful. This combination technique improved the average coverage 23% and for the ceiling 40%.

• Journal of the Korean Society for information Management
• /
• v.26 no.1
• /
• pp.81-105
• /
• 2009

In this paper, we propose user contents using behavior and location information on contents on various channels, such as web, IPTV, for contents distribution. With methods to build user and contents profiles, contents using behavior as an implicit user feedback was applied into machine learning procedure for updating user profiles and contents preference. In machine learning procedure, contents-based and collaborative filtering methods were used to analyze user's contents preference. This study proposes contents location information on web sites for final recommendation contents as well. Finally, we refer to a generalized recommender system for personalization. With those methods, more effective and accurate recommendation service can be possible.

• Journal of KIISE:Software and Applications
• /
• v.32 no.7
• /
• pp.581-591
• /
• 2005

Collaborative filtering systems have problems involving sparsity and the provision of recommendations by making correlations between only two users' preferences. These systems recommend items based only on the preferences without taking in to account the contents of the items. As a result, the accuracy of recommendations depends on the data from user-rated items. When users rate items, it can be expected that not all users ran do so earnestly. This brings down the accuracy of recommendations. This paper proposes a collaborative recommendation method for extracting typical group preferences using user-item matrix optimization and user profiles in collaborative tittering systems. The method excludes unproven users by using entropy based on data from user-rated items and groups users into clusters after generating user profiles, and then extracts typical group preferences. The proposed method generates collaborative user profiles by using association word mining to reflect contents as well as preferences of items and groups users into clusters based on the profiles by using the vector space model and the K-means algorithm. To compensate for the shortcoming of providing recommendations using correlations between only two user preferences, the proposed method extracts typical preferences of groups using the entropy theory The typical preferences are extracted by combining user entropies with item preferences. The recommender system using typical group preferences solves the problem caused by recommendations based on preferences rated incorrectly by users and reduces time for retrieving the most similar users in groups.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.43 no.3
• /
• pp.349-360
• /
• 2019

This study analyzes the trends of recommendation services for customized fashion styles in relation to artificial intelligence. To achieve this goal, the study examined filtering technologies of collaborative, content based, and deep-learning as well as analyzed the characteristics of recommendation services in the users' purchasing process. The results of this study showed that the most universal recommendation technology is collaborative filtering. Collaborative filtering was shown to allow intuitive searching of similar fashion styles in the cognition of need stage, and appeared to be useful in comparing prices but not suitable for innovative customers who pursue early trends. Second, content based filtering was shown to utilize body shape as a key personal profile item in order to reduce the possibility of failure when selecting sizes online, which has limits to being able to wear the product beforehand. Third, fashion style recommendations applied with deep-learning intervene with all user processes of buying products online that was also confirmed to penetrate into the creative area of image tag services, virtual reality services, clothes wearing fit evaluation services, and individually customized design services.

• Journal of the Korea Society of Computer and Information
• /
• v.23 no.8
• /
• pp.45-50
• /
• 2018

Many of the current successful commercial recommender systems utilize collaborative filtering techniques. This technique recommends products to the active user based on product preference history of the neighbor users. Those users with similar preferences to the active user are typically named his/her neighbors. Hence, finding neighbors is critical to performance of the system. Although much effort for developing similarity measures has been devoted in the literature, there leaves a lot to be improved, especially in the aspect of handling subjectivity or vagueness in user preference ratings. This paper addresses this problem and presents a novel similarity measure using fuzzy logic for selecting neighbors. Experimental studies are conducted to reveal that the proposed measure achieved significant performance improvement.