A recommender system for political information filtering

Abstract

Recommender system has been widely used and showcased its successful stories in e-business area for the last decade. It assists in making profits within a lot of companies by recommending their products that the customers would be interested in. Compared to many successful stories in e-business and industries, however, a recommender system has not been fully exploited in non-profit activities where people need information that is unbiased, accurate, up-to-date and mostly relevant to their interest, especially in politics. Even though choosing a right candidate with appropriate and accurate information is required to voters, it is not easy for them to keep up with the political issues due to the massive amounts of online media and its speed. To address these issues, we suggest a politician recommender system by using two widely used filtering: collaborative filtering and content-based filtering. In order to build the recommendation system, we first collect public profile of current congress members in Korea and people’s preference ratings to these politicians. These data are preprocessed and used in filtering methods to recommend politicians that a user would be favorable for. We compare the experimental results, and combine the two filtering whether the hybrid approach shows better performance than two individual methods. We anticipate this saves people’s time and effort to obtain information to support their decision and makes people actively participate in political issues.