Enhancing Movie Recommender Systems with Topic Modeling

• Main Authors: Yen, Ming - Po, 顏銘伯
• Other Authors: Chen, Tsung - Teng
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/pp87t5

碩士 === 國立臺北大學 === 資訊管理研究所 === 105 === In the era of information overloading, the recommender systems may effectively mitigate the information overloading burden and recommend appropriate goods to customers. Since the consumption patterns of products vary, there is no universally applicable recommender system. A recommender needs to be adjusted according to the business situation of the industry. Even in the field of movie recommendation, there are still opportunities to use new data or algorithms to improve the performance. This study applies LDA (Latent Dirichlet Allocation Model) topic modeling technique to analyze the text of the plot of a movie. The resulted topic model serves as the basis for introducing the movies’ plot, director, and other factors to improve the performance of traditional recommendation system. A personalized preference function is then built for each user, which can lower the individual and the overall mean square error to improve the forecast accuracy.