• The KIPS Transactions:PartD
• /
• v.16D no.3
• /
• pp.451-462
• /
• 2009

Collaborative Filtering (CF), one of the most successful technologies among recommender systems, is a system assisting users in easily finding the useful information. However, despite its success and popularity, CF encounters a serious limitation with quality evaluation, called cold start problems. To alleviate this limitation, in this paper, we propose a unique method of building models derived from explicit ratings and applying the models to CF recommender systems. The proposed method is divided into two phases, an offline phase and an online phase. First, the offline phase is a building pre-computed model phase in which most of tasks can be conducted. Second, the online phase is either a prediction or recommendation phase in which the models are used. In a model building phase, we first determine a priori predicted rating and subsequently identify prediction errors for each user. From this error information, an error-reflected model is constructed. The error-reflected model, which is reflected average prior prediction errors of user neighbors and item neighbors, can make accurate predictions in the situation where users or items have few opinions; this is known as the cold start problems. In addition, in order to reduce the re-building tasks, the error-reflected model is designed such that the model is updated effectively and users'new opinions are reflected incrementally, even when users present a new rating feedback.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.406-409
• /
• 2023

With the explosion of information, recommendation algorithms are becoming increasingly important in providing people with appropriate content, enhancing their online experience. In this paper, we propose a recommender system using advanced deep reinforcement learning(DRL) techniques. This method is more adaptive and integrative than traditional methods. We selected the MovieLens dataset and employed the precision metric to assess the effectiveness of our algorithm. The result of our implementation outperforms other baseline techniques, delivering better results for Top-N item recommendations.

• Journal of Intelligence and Information Systems
• /
• v.11 no.1
• /
• pp.107-121
• /
• 2005

Recently, many researches on recommendation systems and collaborative filtering have been proceeding in both research and practice. However, although product items may have multi-valued attributes, previous studies did not reflect the multi-valued attributes. To overcome this limitation, this paper proposes new methodology for recommendation system. The proposed methodology uses multi-valued attributes based on clustering technique for items and applies the collaborative filtering to provide accurate recommendations. In the proposed methodology, both user clustering-based CBR and item attribute clustering-based CBR technique have been applied to the collaborative filtering to consider correlation of item to item as well as correlation of user to user. By using multi-valued attribute-based clustering technique for items, characteristics of items are identified clearly. Extensive experiments have been performed with MovieLens data to validate the proposed methodology. The results of the experiment show that the proposed methodology outperforms the benchmarked methodologies: Case Based Reasoning Collaborative Filtering (CBR_CF) and User Clustering Case Based Reasoning Collaborative Filtering (UC_CBR_CF).

• Journal of Intelligence and Information Systems
• /
• v.23 no.1
• /
• pp.127-141
• /
• 2017

Advances in Internet technologies and the proliferation of mobile devices enabled consumers to approach a wide range of goods and services, while causing an adverse effect that they have hard time reaching their congenial items even if they devote much time to searching for them. Accordingly, businesses are using the recommender systems to provide tools for consumers to find the desired items more easily. Association Rule Mining (ARM) technology is advantageous to recommender systems in that ARM provides intuitive form of a rule with interestingness measures (support, confidence, and lift) describing the relationship between items. Given an item, its relevant items can be distinguished with the help of the measures that show the strength of relationship between items. Based on the strength, the most pertinent items can be chosen among other items and exposed to a given item's web page. However, the diversity of the measures may confuse which items are more recommendable. Given two rules, for example, one rule's support and confidence may not be concurrently superior to the other rule's. Such discrepancy of the measures in distinguishing one rule's superiority from other rules may cause difficulty in selecting proper items for recommendation. In addition, in an online environment where a web page or mobile screen can provide a limited number of recommendations that attract consumer interest, the prudent selection of items to be included in the list of recommendations is very important. The exposure of items of little interest may lead consumers to ignore the recommendations. Then, such consumers will possibly not pay attention to other forms of marketing activities. Therefore, the measures should be aligned with the probability of consumer's acceptance of recommendations. For this reason, this study proposes a model-based approach to combine those measures into one unified measure that can consistently determine the ranking of recommended items. A regression model was designed to describe how well the measures (independent variables; i.e., support, confidence, and lift) explain consumer's acceptance of recommendations (dependent variables, hit rate of recommended items). The model is intuitive to understand and easy to use in that the equation consists of the commonly used measures for ARM and can be used in the estimation of hit rates. The experiment using transaction data from one of the Korea's largest online shopping malls was conducted to show that the proposed model can improve the hit rates of recommendations. From the top of the list to 13th place, recommended items in the higher rakings from the proposed model show the higher hit rates than those from the competitive model's. The result shows that the proposed model's performance is superior to the competitive model's in online recommendation environment. In a web page, consumers are provided around ten recommendations with which the proposed model outperforms. Moreover, a mobile device cannot expose many items simultaneously due to its limited screen size. Therefore, the result shows that the newly devised recommendation technique is suitable for the mobile recommender systems. While this study has been conducted to cover the cross-selling in online shopping malls that handle merchandise, the proposed method can be expected to be applied in various situations under which association rules apply. For example, this model can be applied to medical diagnostic systems that predict candidate diseases from a patient's symptoms. To increase the efficiency of the model, additional variables will need to be considered for the elaboration of the model in future studies. For example, price can be a good candidate for an explanatory variable because it has a major impact on consumer purchase decisions. If the prices of recommended items are much higher than the items in which a consumer is interested, the consumer may hesitate to accept the recommendations.

• Journal of KIISE:Computing Practices and Letters
• /
• v.15 no.3
• /
• pp.189-199
• /
• 2009

The amount of information on the Web is explosively growing as the Internet gains in popularity. However, only a small portion of the information on the Web is truly relevant or useful to the user. Thus, offering suitable information according to user demand is an important subject in information retrieval. In e-commerce, the recommender system is essential to revitalize commercial transactions, raise user satisfaction and loyalty towards the information provider. The existing recommender systems are mostly based on user data collected at servers, so user data are dispersed over several servers. Therefore, web servers that lack sufficient user behavior data cannot easily infer user preferences. Also, if the user visits the server infrequently, it may be hard to reflect the dynamically changing user's interest. This paper proposes a novel personalization system analyzing the user preference based on web documents that are accessed by the user on a user device. The system also identifies non-content blocks appearing repeatedly in the dynamically generated web documents, and adds weight to the keywords extracted from the hyperlink sentence selected by the user. Therefore, the system establishes at an early stage recommendation strategies for the web server that has little user data. Also, user profiles are generated rapidly and more accurately by identifying the information blocks. In order to evaluate the proposed system, this study collected web data and purchase history from users who have current purchase activity. Then, we computed the similarity between purchase data and the user profile. We confirm the accuracy of the generated user profile since the web page containing the purchased item has higher correlation than other item pages.

• The Journal of the Korea Contents Association
• /
• v.8 no.3
• /
• pp.1-7
• /
• 2008

Due to the development of ubiquitous computing, a wide variety of information is being produced and distributed rapidly in digital form. In this excess of information, it is not easy for users to search and find their desired information in short time. In this paper, we propose the personalized item recommendation using the image based filtering. This research uses the image based filtering which is extracting the feature from the image data that a user is interested in, in order to improve the superficial problem of content analysis. We evaluate the performance of the proposed method and it is compared with the performance of previous studies of the content based filtering and the collaborative filtering in the MovieLens dataset. And the results have shown that the proposed method significantly outperforms the previous methods.

• Journal of Intelligence and Information Systems
• /
• v.20 no.4
• /
• pp.107-120
• /
• 2014

This study proposes a novel recommender system using the structural hole analysis to reflect qualitative and emotional information in recommendation process. Although collaborative filtering (CF) is known as the most popular recommendation algorithm, it has some limitations including scalability and sparsity problems. The scalability problem arises when the volume of users and items become quite large. It means that CF cannot scale up due to large computation time for finding neighbors from the user-item matrix as the number of users and items increases in real-world e-commerce sites. Sparsity is a common problem of most recommender systems due to the fact that users generally evaluate only a small portion of the whole items. In addition, the cold-start problem is the special case of the sparsity problem when users or items newly added to the system with no ratings at all. When the user's preference evaluation data is sparse, two users or items are unlikely to have common ratings, and finally, CF will predict ratings using a very limited number of similar users. Moreover, it may produces biased recommendations because similarity weights may be estimated using only a small portion of rating data. In this study, we suggest a novel limitation of the conventional CF. The limitation is that CF does not consider qualitative and emotional information about users in the recommendation process because it only utilizes user's preference scores of the user-item matrix. To address this novel limitation, this study proposes cluster-indexing CF model with the structural hole analysis for recommendations. In general, the structural hole means a location which connects two separate actors without any redundant connections in the network. The actor who occupies the structural hole can easily access to non-redundant, various and fresh information. Therefore, the actor who occupies the structural hole may be a important person in the focal network and he or she may be the representative person in the focal subgroup in the network. Thus, his or her characteristics may represent the general characteristics of the users in the focal subgroup. In this sense, we can distinguish friends and strangers of the focal user utilizing the structural hole analysis. This study uses the structural hole analysis to select structural holes in subgroups as an initial seeds for a cluster analysis. First, we gather data about users' preference ratings for items and their social network information. For gathering research data, we develop a data collection system. Then, we perform structural hole analysis and find structural holes of social network. Next, we use these structural holes as cluster centroids for the clustering algorithm. Finally, this study makes recommendations using CF within user's cluster, and compare the recommendation performances of comparative models. For implementing experiments of the proposed model, we composite the experimental results from two experiments. The first experiment is the structural hole analysis. For the first one, this study employs a software package for the analysis of social network data - UCINET version 6. The second one is for performing modified clustering, and CF using the result of the cluster analysis. We develop an experimental system using VBA (Visual Basic for Application) of Microsoft Excel 2007 for the second one. This study designs to analyzing clustering based on a novel similarity measure - Pearson correlation between user preference rating vectors for the modified clustering experiment. In addition, this study uses 'all-but-one' approach for the CF experiment. In order to validate the effectiveness of our proposed model, we apply three comparative types of CF models to the same dataset. The experimental results show that the proposed model outperforms the other comparative models. In especial, the proposed model significantly performs better than two comparative modes with the cluster analysis from the statistical significance test. However, the difference between the proposed model and the naive model does not have statistical significance.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.21 no.6
• /
• pp.1183-1190
• /
• 2017

With the development of network and IT technology, people are searching and purchasing items they want, not bounded by places. Therefore, there are various studies on how to solve the scalability problem due to the rapidly increasing data in the recommendation system. In this paper, we propose an item-based collaborative filtering method using Tag weight and a recommendation technique using MapReduce method, which is a distributed parallel processing method. In order to improve speed and efficiency, the proposed method classifies items into categories in the preprocessing and groups according to the number of nodes. In each distributed node, data is processed by going through Map-Reduce step 4 times. In order to recommend better items to users, item tag weight is used in the similarity calculation. The experiment result indicated that the proposed method has been more enhanced the appropriacy compared to item-based method, and run efficiently on the large amounts of data.

• Journal of the Korea Society of Computer and Information
• /
• v.18 no.10
• /
• pp.23-37
• /
• 2013

In this paper, the proposed movie recommender system constructs trust network, which is similar to social network, using user's trust information that users explicitly present. Recommendation on items is performed by using relation degree between users and information of recommended item is provided by a visualization method. We discover the hidden relationships via the constructed trust network. To provide visualized recommendation information, we employ Freebase which is large knowledge base supporting information such as movie, music, and people in structured format. We provide three visualization methods as the followings: i) visualization based on movie posters with the number of movies that user required. ii) visualization on extra information such as director, actor and genre and so on when user selected a movie from recommendation list. iii) visualization based on movie posters that is recommended by neighbors who a user selects from trust network. The proposed system considers user's social relations and provides visualization which can reflect user's requirements. Using the visualization methods, user can reach right decision making on items. Furthermore, the proposed system reflects the user's opinion through recommendation visualization methods and can provide rich information to users through LOD(Linked Open Data) Cloud such as Freebase, LinkedMDB and Wikipedia and so on.

• Journal of Intelligence and Information Systems
• /
• v.14 no.2
• /
• pp.157-177
• /
• 2008

Even in a single day, an enormous amount of content including digital videos, posts, photographs, and wikis are generated on the web. It's getting more difficult to recommend to a user what he/she prefers among these contents because of the difficulty of automatically grasping of content's meanings. CF (Collaborative Filtering) is one of useful methods to recommend proper content to a user under these situations because the filtering process is only based on historical information about whether or not a target user has preferred an item before. Collaborative Tagging is the process that allows many users to annotate content with descriptive tags. Recommendation using tags can partially improve, such as the limitations of CF, the sparsity and cold-start problem. In this research, a CF method with user-created tags is proposed. Collaborative tagging is employed to grasp and filter users' preferences for items. Empirical demonstrations using real dataset from del.icio.us show that our algorithm obtains improved performance, compared with existing works.