Comparison of the Constant Prediction Time of Collaborative Filtering Algorithms by Using Time Contexts

This research presents the comparison of collaborative filtering techniques which are Tendencies Based Algorithm, Item mean algorithm, and Simple mean based algorithm. All these algorithms use the constant time in prediction process. To evaluate our proposed model, we use last. fm dataset including music listening history of each user. Each user's profile is split into several sub-profiles based on specified time ranges called "Time Contexts". Thus the prediction is done using these Time Contexts instead of a single user profile. From our experiments, we have found that Tendencies Based Algorithm with Time Contexts is effective. It is given more accuracy and much more efficient computationally than tradition collaborative filtering algorithms.