Optimization of User Based Collaborative Filtering

With the advancement of technologies in the modern era, the amount of data that emerges from various sources like online websites, Internet of things, E-commerce etc., keeps on increasing at a larger pace. The data available is too much for a common user to handle. So recommendation system makes efforts to provide right information to the right user at their doorstep and makes it easy for the users. Similarity measure is considered as an important step to determine the accuracy of the recommendation system. A classical collaborative filtering is implemented either by using Pearson correlation coefficient or Cosine similarity which has got its own merits and shortcomings. We propose an enhanced similarity measure by applying the set based methodology on basic similarity measures and analyze the impact of those various enhanced similarity measures such as set based cosine, set based Pearson correlation coefficient, set based Spearman, set based Kendall on the user based collaborative filtering recommendation systems. It was observed that the enhanced similarity measure obtained from set based methodologies were more significant than the basic measures using Wilcoxon test.