• Journal of Intelligence and Information Systems
• /
• v.26 no.2
• /
• pp.43-56
• /
• 2020

Recommender systems help users make the best choice among various options. Especially, recommender systems play important roles in internet sites as digital information is generated innumerable every second. Many studies on recommender systems have focused on an accurate recommendation. However, there are some problems to overcome in order for the recommendation system to be commercially successful. First, there is a lack of transparency in the recommender system. That is, users cannot know why products are recommended. Second, the recommender system cannot immediately reflect changes in user preferences. That is, although the preference of the user's product changes over time, the recommender system must rebuild the model to reflect the user's preference. Therefore, in this study, we proposed a recommendation methodology using topic modeling and sequential association rule mining to solve these problems from review data. Product reviews provide useful information for recommendations because product reviews include not only rating of the product but also various contents such as user experiences and emotional state. So, reviews imply user preference for the product. So, topic modeling is useful for explaining why items are recommended to users. In addition, sequential association rule mining is useful for identifying changes in user preferences. The proposed methodology is largely divided into two phases. The first phase is to create user profile based on topic modeling. After extracting topics from user reviews on products, user profile on topics is created. The second phase is to recommend products using sequential rules that appear in buying behaviors of users as time passes. The buying behaviors are derived from a change in the topic of each user. A collaborative filtering-based recommendation system was developed as a benchmark system, and we compared the performance of the proposed methodology with that of the collaborative filtering-based recommendation system using Amazon's review dataset. As evaluation metrics, accuracy, recall, precision, and F1 were used. For topic modeling, collapsed Gibbs sampling was conducted. And we extracted 15 topics. Looking at the main topics, topic 1, top 3, topic 4, topic 7, topic 9, topic 13, topic 14 are related to "comedy shows", "high-teen drama series", "crime investigation drama", "horror theme", "British drama", "medical drama", "science fiction drama", respectively. As a result of comparative analysis, the proposed methodology outperformed the collaborative filtering-based recommendation system. From the results, we found that the time just prior to the recommendation was very important for inferring changes in user preference. Therefore, the proposed methodology not only can secure the transparency of the recommender system but also can reflect the user's preferences that change over time. However, the proposed methodology has some limitations. The proposed methodology cannot recommend product elaborately if the number of products included in the topic is large. In addition, the number of sequential patterns is small because the number of topics is too small. Therefore, future research needs to consider these limitations.

• The KIPS Transactions:PartB
• /
• v.19B no.3
• /
• pp.155-164
• /
• 2012

This study aims to identify the multidimensional variables and sub-variables and study their relative weight in music recommender systems when maximizing the rating function R. To undertake the task, a optimization formula and variables for a research model were derived from the review of prior works on recommender systems, which were then used to establish the research model for an empirical test. With the research model and the actual log data of real customers obtained from an on line music provider in Korea, multiple regression analysis was conducted to induce the optimal correlation of variables in the multidimensional model. The results showed that the correlation value against the rating function R for Items was highest, followed by Social Relations, Users and Contexts. Among sub-variables, popular music from Social Relations, genre, latest music and favourite artist from Items were high in the correlation with the rating function R. Meantime, the derived multidimensional recommender systems revealed that in a comparative analysis, it outperformed two dimensions(Users, Items) and three dimensions(Users, Items and Contexts, or Users, items and Social Relations) based recommender systems in terms of adjusted $R^2$ and the correlation of all variables against the values of the rating function R.

• Journal of Intelligence and Information Systems
• /
• v.26 no.1
• /
• pp.151-166
• /
• 2020

A recommender system is a system that recommends products or services that best meet the preferences of each customer using statistical or machine learning techniques. Collaborative filtering (CF) is the most commonly used algorithm for implementing recommender systems. However, in most cases, it only uses purchase history or customer ratings, even though customers provide numerous other data that are available. E-commerce customers frequently use a search function to find the products in which they are interested among the vast array of products offered. Such search keyword data may be a very useful information source for modeling customer preferences. However, it is rarely used as a source of information for recommendation systems. In this paper, we propose a novel hybrid CF model based on the Doc2Vec algorithm using search keywords and purchase history data of online shopping mall customers. To validate the applicability of the proposed model, we empirically tested its performance using real-world online shopping mall data from Korea. As the number of recommended products increases, the recommendation performance of the proposed CF (or, hybrid CF based on the customer's search keywords) is improved. On the other hand, the performance of a conventional CF gradually decreased as the number of recommended products increased. As a result, we found that using search keyword data effectively represents customer preferences and might contribute to an improvement in conventional CF recommender systems.

• Knowledge Management Research
• /
• v.21 no.3
• /
• pp.27-44
• /
• 2020

In order to solve the user's information overload problem, recommender systems infer users' preferences and suggest items that match them. The collaborative filtering (CF), the most successful recommendation algorithm, has been improving performance until recently and applied to various business domains. Visual information, such as book covers, could influence consumers' purchase decision making. However, CF-based recommender systems have rarely considered for visual information. In this study, we propose VizNCS, a CF-based deep learning model that uses visual information as additional information. VizNCS consists of two phases. In the first phase, we build convolutional neural networks (CNN) to extract visual features from image data. In the second phase, we supply the visual features to the NCF model that is known to easy to extend to other information among the deep learning-based recommendation systems. As the results of the performance comparison experiments, VizNCS showed higher performance than the vanilla NCF. We also conducted an additional experiment to see if the visual information affects differently depending on the product category. The result enables us to identify which categories were affected and which were not. We expect VizNCS to improve the recommender system performance and expand the recommender system's data source to visual information.

• Information Systems Review
• /
• v.18 no.3
• /
• pp.155-183
• /
• 2016

Research on personalized recommender service that uses big data has gained considerable attention given the increasing volume of contents being created. This development indicates the need for service providers to collect personal information and content rating data to personalize content recommendations. Previous studies on this topic proposed algorithms to offer improved recommendations using minimal rating data or service designs and increase the number of ratings. However, limited studies have been conducted on the factors that motivate the ratings input of users, as well as the factors that influence their continuous usage of recommender service. The present study explored the factors that motivate users to enter ratings by conducting in-depth interviews with users who use recommender services. The meanings of these ratings were also explored. Results show that the meaning and usage range of ratings differed based on the stage of a user's with utilization of the service. When users input an initial rating, they treat such a rating as a database to save the impression of a past experience. Such a rating is then used as a tool to reflect the current feeling and thoughts of a user. In the end, users were not only interested in their own rating system, but they also actively sought out the meaning of the rating systems of others and utilized them. Users also expressed mistrust in the recommendations of the service because they were aware of the limitation of the algorithms. This study identified a number of practical implications regarding recommender services.

• Information Systems Review
• /
• v.19 no.2
• /
• pp.71-94
• /
• 2017

Recommender systems reduce information overload and enhance choice quality. This technology is used in many services and industry. Previous studies did not consider recommendation quantity and the repetitive recommendations of an item. This study is the first to examine recommender systems by considering recommendation quantity and repetitive recommendations. Only a limited number of items are displayed in offline stores because of their physical limitations. Determining the type and number of items that will be displayed is an important consideration. In this study, I suggest the use of a user-based recommender system that can recommend the most appropriate items for each store. This model is evaluated by MAE, Precision, Recall, and F1 measure, and shows higher performance than the baseline model. I also suggest a new performance evaluation measure that includes Quantity Precision, Quantity Recall, and Quantity F1 measure. This measure considers the penalty for short or excess recommendation quantity. Novelty is defined as the proportion of items in a recommendation list that consumers may not experience. I evaluate the new revenue creation effect of the suggested model using this novelty measure. Previous research focused on recommendations for customer online, but I expand the recommender system to cover stores offline.

• Journal of the Korean Data and Information Science Society
• /
• v.17 no.3
• /
• pp.717-726
• /
• 2006

This study is to investigate the MAE of prediction value by collaborative filtering algorithm originated by GroupLens and improved algorithm. To decrease the MAE on the collaborative recommender system on user based, this research proposes the improved algorithm, which reduces the possibility of over estimation of active user's preference mean collaboratively using other user’s preference mean. The result shows the MAE of prediction by improved algorithm is better than original algorithm, so the active user's preference mean used in prediction formula is possibly over estimated.

• Journal of the Korea Society of Computer and Information
• /
• v.25 no.9
• /
• pp.101-108
• /
• 2020

Recommender system is a must-have feature of e-commerce, since it provides customers with convenience in selecting products. Collaborative filtering is a widely-used and representative technique, where it gives recommendation lists of products preferred by other users or preferred by the current user in the past. Recently, researches on the recommendation system using deep learning artificial intelligence technologies are actively being conducted to achieve performance improvement. This study develops a collaborative filtering based recommender system using restricted Boltzmann machines of the deep learning technology by utilizing user ratings. Moreover, a learning parameter update algorithm is proposed for learning efficiency and performance. Performance evaluation of the proposed system is made through experimental analysis and comparison with conventional collaborative filtering methods. It is found that the proposed algorithm yields superior performance than the basic restricted Boltzmann machines.

• Journal of Information Technology Applications and Management
• /
• v.21 no.4
• /
• pp.203-223
• /
• 2014

This study proposes a novel context-aware recommender system, which is designed to recommend the items according to the customer's responses to the previously recommended item. In specific, our proposed system predicts the user's emotional state from his or her responses (such as facial expressions and movements) to the previous recommended item, and then it recommends the items that are similar to the previous one when his or her emotional state is estimated as positive. If the customer's emotional state on the previously recommended item is regarded as negative, the system recommends the items that have characteristics opposite to the previous item. Our proposed system consists of two sub modules-(1) emotion prediction module, and (2) responsive recommendation module. Emotion prediction module contains the emotion prediction model that predicts a customer's arousal level-a physiological and psychological state of being awake or reactive to stimuli-using the customer's reaction data including facial expressions and body movements, which can be measured using Microsoft's Kinect Sensor. Responsive recommendation module generates a recommendation list by using the results from the first module-emotion prediction module. If a customer shows a high level of arousal on the previously recommended item, the module recommends the items that are most similar to the previous item. Otherwise, it recommends the items that are most dissimilar to the previous one. In order to validate the performance and usefulness of the proposed recommender system, we conducted empirical validation. In total, 30 undergraduate students participated in the experiment. We used 100 trailers of Korean movies that had been released from 2009 to 2012 as the items for recommendation. For the experiment, we manually constructed Korean movie trailer DB which contains the fields such as release date, genre, director, writer, and actors. In order to check if the recommendation using customers' responses outperforms the recommendation using their demographic information, we compared them. The performance of the recommendation was measured using two metrics-satisfaction and arousal levels. Experimental results showed that the recommendation using customers' responses (i.e. our proposed system) outperformed the recommendation using their demographic information with statistical significance.

• Journal of KIISE:Information Networking
• /
• v.36 no.1
• /
• pp.50-56
• /
• 2009

This paper proposes a new algorithm of clustering similar nodes defined as nodes having similar characteristic values in pure P2P environment. To compare similarity between nodes, we introduce a fitness function whose return value depends only on the two nodes' characteristic values. The higher the return value is, the more similar the two nodes are. We propose a GORGFM algorithm newly in conjunction with the fitness function to recommend and exchange nodes' characteristic values for an interest group formation and maintenance. With the GORGFM algorithm, the interest groups are formed dynamically based on the similarity of users, and all nodes will highly satisfy with the information recommended and received from nodes of the interest group. To evaluate of performance of the GORGFM algorithm, we simulated a matching rate by the total number of nodes of network and the number of iterations of the algorithm to find similar nodes accurately. The result shows that the matching rate is highly accurate. The GORGFM algorithm proposed in this paper is highly flexible to be applied for any searching system on the web.