Traditional recommendation systems are mostly based on similarity discrimination which requires sufficient data and recommends high correlated items. It becomes very difficult to accurately recommend products when data are not enough. Thus, the research a