Cellular phone ringing tone recommendation system based on collaborative filtering method
暂无分享,去创建一个
We have developed a prototype of cellular phone ringing tome recommendation system using memory-based collaborative filtering and we have carried out examinations to evaluate its performance. The ringing tone content was stored on a server from where the users were able to download the desired items according to their preferences. An extensive log data accumulated at the download service site for a fixed period of time was used. The log data contained only information for the users' downloaded ringing tomes without evaluation data. The user set and the tone downloadable content set were not fixed and our goal was to investigate how collaborative filtering could be successfully applied to a system with such continuously changing conditions. The Jaccard's similarity coefficient was used to calculate the similarity between the users. The learning period, the recommendation period and the number of the similar users were used as condition parameters. The system quality evaluation showed that the recall increases with the increase of the learning period but decreases with the increase of the recommendation period. Optimal values for the number of the most similar users as well as for the learning and the recommendation periods were experimentally obtained. It was shown that the collaborative filtering method could be successfully applied to a cellular phone ringing tone recommendation system.
[1] David Heckerman,et al. Empirical Analysis of Predictive Algorithms for Collaborative Filtering , 1998, UAI.
[2] John Riedl,et al. GroupLens: an open architecture for collaborative filtering of netnews , 1994, CSCW '94.
[3] Michael J. Pazzani,et al. Improvement of Collaborative Filtering with the Simple Bayesian Classifier 1 , 2002 .
[4] David Maxwell Chickering,et al. A Bayesian Approach to Learning Bayesian Networks with Local Structure , 1997, UAI.