• Information Systems Review
• /
• v.9 no.1
• /
• pp.85-103
• /
• 2007

The purpose of this study is to evaluate the performance of collaborative filtering recommender algorithms for better prediction accuracy of the customer's preference. The accuracy of customer's preference prediction is compared through the MAE of neighborhood based collaborative filtering algorithm and correspondence mean algorithm. It is analyzed by using MovieLens 1 Million dataset in order to experiment with the prediction accuracy of the algorithms. For similarity, weight used in both algorithms, commonly, Pearson's correlation coefficient and vector similarity which are used generally were utilized, and as a result of analysis, we show that the accuracy of the customer's preference prediction of correspondence mean algorithm is superior. Pearson's correlation coefficient and vector similarity used in two algorithms are calculated using the preference rating of two customers' co-rated movies, and it shows that similarity weight is overestimated, where the number of co-rated movies is small. Therefore, it is intended to increase the accuracy of customer's preference prediction through expanding the number of the existing co-rated movies.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.13 no.5
• /
• pp.526-529
• /
• 2003

Recommender systems or collaborative filtering are methods to identify potentially interesting or valuable items to a particular user Under the assumption that people with similar interest tend to like the similar types of items, these methods use a database on the preference of a set of users and predict the rating on the items that the user has not rated. Usually the preference of a particular user is liable to vary with time and this temporal variation may cause an inaccurate identification and prediction. In this paper we propose a method to adapt the temporal variation of the user preference in order to improve the predictive performance of a collaborative filtering algorithm. To be more specific, the correlation weight of the GroupLens system which is a general formulation of statistical collaborative filtering algorithm is modified to reflect only recent similarity between two user. The proposed method is evaluated for EachMovie dataset and shows much better prediction results compared with GrouPLens system.

• The Korean Journal of Applied Statistics
• /
• v.29 no.4
• /
• pp.657-678
• /
• 2016

This study deals with the effect of online word-of-mouth (OWOM) variables on the box office. From the result of statistical analysis on 276 films with audiences of more than five hundred thousand released in the Korea from 2012 to 2015, it can be seen that the variables showing the size of OWOM (such as the number of the portal movie rater, blog, and news after release) are associated more with the box office than the portal movie rating showing the direction of OWOM as well as variables showing the inherent properties of the film such as grade, nationality, release month, release season, directors, actors, and distributors.

• Journal of KIISE:Software and Applications
• /
• v.32 no.11
• /
• pp.1111-1120
• /
• 2005

In collaborative filtering systems most users do not rate preferences; so User-Item matrix shows great sparsity because it has missing values for items not rated by users. Generally, the systems predict the preferences of an active user based on the preferences of a group of users. However, default voting methods predict all missing values for all users in User-Item matrix. One of the most common methods predicting default voting values tried two different approaches using the average rating for a user or using the average rating for an item. However, there is a problem that they did not consider the characteristics of items, users, and the distribution of data set. We replace the missing values in the User-Item matrix by the default noting method using user coefficient of variance. We select the threshold of user coefficient of variance by using equations automatically and determine when to shift between the user averages and item averages according to the threshold. However, there are not always regular relations between the averages and the thresholds of user coefficient of variances in datasets. It is caused that the distribution information of user coefficient of variances in datasets affects the threshold of user coefficient of variance as well as their average. We decide the threshold of user coefficient of valiance by combining them. We evaluate our method on MovieLens dataset of user ratings for movies and show that it outperforms previously default voting methods.

• Journal of KIISE:Computing Practices and Letters
• /
• v.9 no.5
• /
• pp.521-529
• /
• 2003

In this paper we present an improved method by using demographic information for overcoming the similarity miss-calculation from the sparsity problem in collaborative filtering recommendation systems. The similarity between a pair of users is only determined by the ratings given to co-rated items, so items that have not been rated by both users are ignored. To solve this problem, we add virtual neighbor's rating using demographic information of neighbors for improving prediction accuracy. It is one kind of extentions of traditional collaborative filtering methods using the peason correlation coefficient. We used the Grouplens movie rating data in experiment and we have compared the proposed method with the collaborative filtering methods by the mean absolute error and receive operating characteristic values. The results show that the proposed method is more efficient than the collaborative filtering methods using the pearson correlation coefficient about 9% in MAE and 13% in sensitivity of ROC.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.05a
• /
• pp.357-359
• /
• 2023

This paper proposes a hybrid recommendation system (RS) model that overcomes the limitations of traditional approaches such as data sparsity, cold start, and scalability by combining collaborative filtering and context-aware techniques. The objective of this model is to enhance the accuracy of recommendations and provide personalized suggestions by leveraging the strengths of collaborative filtering and incorporating user context features to capture their preferences and behavior more effectively. The approach utilizes a novel method that combines contextual attributes with the original user-item rating matrix of CF-based algorithms. Furthermore, we integrate k-mean++ clustering to group users with similar preferences and finally recommend items that have highly rated by other users in the same cluster. The process of partitioning is the use of the rating matrix into clusters based on contextual information offers several advantages. First, it bypasses of the computations over the entire data, reducing runtime and improving scalability. Second, the partitioned clusters hold similar ratings, which can produce greater impacts on each other, leading to more accurate recommendations and providing flexibility in the clustering process. keywords: Context-aware Recommendation, Collaborative Filtering, Kmean++ Clustering.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.219-239
• /
• 2019

As the importance of providing customized services to individuals becomes important, researches on personalized recommendation systems are constantly being carried out. Collaborative filtering is one of the most popular systems in academia and industry. However, there exists limitation in a sense that recommendations were mostly based on quantitative information such as users' ratings, which made the accuracy be lowered. To solve these problems, many studies have been actively attempted to improve the performance of the recommendation system by using other information besides the quantitative information. Good examples are the usages of the sentiment analysis on customer review text data. Nevertheless, the existing research has not directly combined the results of the sentiment analysis and quantitative rating scores in the recommendation system. Therefore, this study aims to reflect the sentiments shown in the reviews into the rating scores. In other words, we propose a new algorithm that can directly convert the user 's own review into the empirically quantitative information and reflect it directly to the recommendation system. To do this, we needed to quantify users' reviews, which were originally qualitative information. In this study, sentiment score was calculated through sentiment analysis technique of text mining. The data was targeted for movie review. Based on the data, a domain specific sentiment dictionary is constructed for the movie reviews. Regression analysis was used as a method to construct sentiment dictionary. Each positive / negative dictionary was constructed using Lasso regression, Ridge regression, and ElasticNet methods. Based on this constructed sentiment dictionary, the accuracy was verified through confusion matrix. The accuracy of the Lasso based dictionary was 70%, the accuracy of the Ridge based dictionary was 79%, and that of the ElasticNet (${\alpha}=0.3$) was 83%. Therefore, in this study, the sentiment score of the review is calculated based on the dictionary of the ElasticNet method. It was combined with a rating to create a new rating. In this paper, we show that the collaborative filtering that reflects sentiment scores of user review is superior to the traditional method that only considers the existing rating. In order to show that the proposed algorithm is based on memory-based user collaboration filtering, item-based collaborative filtering and model based matrix factorization SVD, and SVD ++. Based on the above algorithm, the mean absolute error (MAE) and the root mean square error (RMSE) are calculated to evaluate the recommendation system with a score that combines sentiment scores with a system that only considers scores. When the evaluation index was MAE, it was improved by 0.059 for UBCF, 0.0862 for IBCF, 0.1012 for SVD and 0.188 for SVD ++. When the evaluation index is RMSE, UBCF is 0.0431, IBCF is 0.0882, SVD is 0.1103, and SVD ++ is 0.1756. As a result, it can be seen that the prediction performance of the evaluation point reflecting the sentiment score proposed in this paper is superior to that of the conventional evaluation method. In other words, in this paper, it is confirmed that the collaborative filtering that reflects the sentiment score of the user review shows superior accuracy as compared with the conventional type of collaborative filtering that only considers the quantitative score. We then attempted paired t-test validation to ensure that the proposed model was a better approach and concluded that the proposed model is better. In this study, to overcome limitations of previous researches that judge user's sentiment only by quantitative rating score, the review was numerically calculated and a user's opinion was more refined and considered into the recommendation system to improve the accuracy. The findings of this study have managerial implications to recommendation system developers who need to consider both quantitative information and qualitative information it is expect. The way of constructing the combined system in this paper might be directly used by the developers.

• Journal of Distribution Science
• /
• v.15 no.10
• /
• pp.41-50
• /
• 2017

Purpose - Online product review has become a crucial part of the online retailer's market performance for a wide range of products. This research aims to investigate how an individual reviewer's review frequency and timing affect her/his average attitude toward products. Research design, data, and methodology - To conduct reviewer-level analysis, this study uses 42,172 posted online review messages generated by 6,941 identified reviewers for 59 movies released in the South Korea from July 2015 to December 2015. This study adopts Tobit model specification to take into account the censored nature and the selection bias arising from the nature of J-shaped distribution of movie rating. Results - Our estimation results support that the negative impact of review frequency and timing on valence. Furthermore, review timing has an inverted-U relationship with the user's average valence and enhance the negative effect of review frequency. Conclusions - This study contributes to the growing literature on the understanding how eWOM is generated at the individual consumer level. On the basis of the main empirical findings, this study provides insights into building a recommendation system in online retail store based on the consumer's review history data - frequency, timing, and valence.

• Journal of Intelligence and Information Systems
• /
• v.22 no.2
• /
• pp.97-113
• /
• 2016

Due to the growth of internet data and the rapid development of internet technology, "big data" analysis is actively conducted to analyze enormous data for various purposes. Especially in recent years, a number of studies have been performed on the applications of text mining techniques in order to overcome the limitations of existing structured data analysis. Various studies on sentiment analysis, the part of text mining techniques, are actively studied to score opinions based on the distribution of polarity of words in documents. Usually, the sentiment analysis uses sentiment dictionary contains positivity and negativity of vocabularies. As a part of such studies, this study tries to construct sentiment dictionary which is customized to specific data domain. Using a common sentiment dictionary for sentiment analysis without considering data domain characteristic cannot reflect contextual expression only used in the specific data domain. So, we can expect using a modified sentiment dictionary customized to data domain can lead the improvement of sentiment analysis efficiency. Therefore, this study aims to suggest a way to construct customized dictionary to reflect characteristics of data domain. Especially, in this study, movie review data are divided by genre and construct genre-customized dictionaries. The performance of customized dictionary in sentiment analysis is compared with a common sentiment dictionary. In this study, IMDb data are chosen as the subject of analysis, and movie reviews are categorized by genre. Six genres in IMDb, 'action', 'animation', 'comedy', 'drama', 'horror', and 'sci-fi' are selected. Five highest ranking movies and five lowest ranking movies per genre are selected as training data set and two years' movie data from 2012 September 2012 to June 2014 are collected as test data set. Using SO-PMI (Semantic Orientation from Point-wise Mutual Information) technique, we build customized sentiment dictionary per genre and compare prediction accuracy on review rating. As a result of the analysis, the prediction using customized dictionaries improves prediction accuracy. The performance improvement is 2.82% in overall and is statistical significant. Especially, the customized dictionary on 'sci-fi' leads the highest accuracy improvement among six genres. Even though this study shows the usefulness of customized dictionaries in sentiment analysis, further studies are required to generalize the results. In this study, we only consider adjectives as additional terms in customized sentiment dictionary. Other part of text such as verb and adverb can be considered to improve sentiment analysis performance. Also, we need to apply customized sentiment dictionary to other domain such as product reviews.

• Journal of Internet Computing and Services
• /
• v.12 no.2
• /
• pp.85-101
• /
• 2011

The recommender system analyzes users' preference and predicts the users' preference to items in order to recommend various items such as book, movie and music for the users. The collaborative filtering method is used most widely in the recommender system. The method uses rating information of similar users when recommending items for the target users. Performance of the collaborative filtering-based recommendation is lowered when attacker maliciously manipulates the rating information on items. This kind of malicious act on a recommender system is called 'Recommendation Attack'. When the evaluation data that are in continuous change are analyzed in the perspective of data stream, it is possible to predict attack on the recommender system. In this paper, we will suggest the method to detect attack on the recommender system by using the stream trend of the item evaluation in the collaborative filtering-based recommender system. Since the information on item evaluation included in the evaluation data tends to change frequently according to passage of time, the measurement of changes in item evaluation in a fixed period of time can enable detection of attack on the recommender system. The method suggested in this paper is to compare the evaluation stream that is entered continuously with the normal stream trend in the test cycle for attack detection with a view to detecting the abnormal stream trend. The proposed method can enhance operability of the recommender system and re-usability of the evaluation data. The effectiveness of the method was verified in various experiments.