Evaluation of Collaborative Filtering Techniques for Mobile and Web Application Based Recommendation System

In this paper, performance evaluation of collaborative filtering approaches has been presented, which is to be used in a mobile application and web based recommendation system that is under development for indoor decoration. Recommendation system consists of two parts. The image captured by a mobile phone is analyzed using color quantization methods, and results of the analysis are transferred to the web section. The study in this paper deals with only recommendation part of this system, in which we evaluate the performance of various collaborative filtering methods such as correlation threshold and K-nearest neighboring. We have used Movilens data sets which present scores voted by users for different movies. Movilens data sets are similar to data sets to be used in our recommendation system. Results show the feasibility of using K-nearest neigboring method with Pearson similarity.