Abstract
Recommender System (RS) is an information filtering software tool that provides relevant recommendations to users on various products. Finding the preference of users on products is the core component of RS. Most of the available RS datasets are complex enough to represent the user preferences and usually require a lot of processing before their utilization. On the other hand, selecting an appropriate learning algorithm for making relevant recommendations is a complex task. In this paper, we perform various pre-processing steps on a movie dataset to find the key features that help us to identify the actual user preferences. Moreover, we perform a comparative analysis of various randomized and non-randomized learning algorithms by utilizing these features to identify the best learning algorithm. Our study reinforces the superiority of randomized learning algorithms over non-randomized ones on MovieLens dataset.