Integrating new community cold start for recommender systems

Recommender Systems (or RS) are of great importance to Web 2.0 applications. New community cold start is remaining a challenging work for RS, due to there exists little users holding similar preferences with the new community cold start users. The relevant existing Collaborative Filtering (CF)-based RS fall short in dealing with the new community cold start problem. In this paper, a Degree of Item Correlation (DIC)-based Scale and Translation Invariant (or DIC-STI) approach is proposed to improve the recommendation accuracy for CF-based RS in the situation of new community cold start. Moreover, an integrated recommendation approach supporting new community cold start (using DIC-STI) is also presented based on the traditional RS. Experimental results demonstrate the higher recommendation accuracy than existing methodologies. Copyright © 2015 Binary Information Press.