Improving the Pearson Similarity Equation for Recommender Systems by Age Parameter

Abstract

Recent years, collaborative filtering systems have become so much popular due to excessive developments in technology. Rooting from the machine learning, recommender systems are in the center of the knowledge engineering that are used frequently in many computer science related fields. Therefore, this paper tries to utilize time information inferred from users' ages, which is going to be employed in the recommender systems during the similarity calculation between users. Adding more parameters, in other words, inserting detailed information of the users' preferences into the calculations will stimulate the resolution of fulfilling results thanks to new criterion. Specifically, it is proposed a conversion between two particular dates which are relevant to the user's personal data. 10-year interval is considered as a transition between generations, thus generation gap between users can make sense positively or negatively in terms of the amount of gap. Additionally, for the prerequisite of being a peer, it is defined a 3-year long difference can be a boundary to say that two users can be in the same wavelength. From the inspiration of this idea, a new mathematical addition into the well-known similarity calculation is presented in this paper. Eventually, this further time detail is taken into account via inverse proportion in similarity calculation and time incorporation approach for recommender system is emphasized. All in all, proposed method is to be tested with a real valued Movie Lens-MLP dataset on the platform of MatLab.