New Recommendation System Model Based on Semantic Similarity in Movie Domain
Amir Hossein, Nabizadeh Rafsanjani, Naomie Binti Salim, Nastaran Mohammadhossein, Karamollah Bagheri Fard · 2013
Recommender systems (RS) automatically select the most appropriate items to each user, thus shortening his product searching time and adapting the selection as his particular preferences evolve over time. It has been found that different recommendation methods use different techniques to recommend objects to consumers. There is possibility that the majority of the previous researches have problems. Currently, the most important problem in recommendation system is cold start that during this study, we consider to this issue by proposing RECOMOVIER model. Four techniques have been developed for designing new model including: Collaborative Filtering (CF), Content Base Filtering (CBF), and Hybrid and Cascade method. The main goal of this research is omitting the cold start problem for new user that helps to increase the accuracy of results of recommendation system. Analytical review of existing recommendation system model is utilized as research methodology which guides us to understand more about pros and cons of current recommender methods and systems. In order to implement the proposed model the C# programming language and two sets of data that were downloaded and modified from Movielens website were used. This application is tested by users as every user filled the form and then got feedback from this application. The results of experiments show that, using user profile has positive influence on increasing the accuracy of results of recommendation system by omitting the cold start problem from RS systems. The proposed model in this paper would be valuable and beneficial for future researchers and practitioners interested in developing recommendation systems. KEYWORDS: Recommendation Systems (RS), Collaborative Filtering (CF), Content Base Filtering (CBF), Hybrid