MRBUP : MapReduce based scalable user profiling in recommender system

Abstract

The speed at which data is generated has accelerated over the years and is predicted to grow exponentially in the future. It is preferable to pique the user’s interest by building a user profile, then filter the products based on their preferences. User profiling becomes necessary as it enables the system to understand the user’s tastes and likes from voluminous data. To build a scalable system, this paper proposes MRBUP which focuses on creating scalable user profile by using the model of MapReduce. Users in a recommender system interacts with products, which are humungous in nature. To build user profile, MRBUP practices term frequency-inverse document frequency technique. For experimental analysis, this work uses the Amazon Product Reviews dataset which defines product reviews given by users to different products in the catalog. MRBUP utilizes the product reviews given by the user to the products which are in pure textual form to generate user profile. MRBUP is executed on a Hadoop cluster of varying size and computational time is observed. Scalability MRBUP is evaluated using the usual metric of Speedup factor.