High-Scale Food Recommendation Built on Apache Spark using Alternating Least Squares

Abstract

Recommendation system is an information filtering application which anticipates the user's ratings by considering their previous judgements. These systems are an extensive aspect of most state-of-the-art applications. The highly accessible data on the internet is playing a dominant role in these systems. Collaborative filtering is one of the mechanisms of building a recommendation system which when fabricated with the ALS algorithm promises stunning results. This is because of the two-step iterative matrix factorisation approach of ALS. This recommendation system has been built using Amazon food product reviews on Apache Spark by servicing a master and few slave nodes. PySpark's ml library has been used for constructing the ALS model and RDD were utilised for handling the colossal amount of data. Composing the model over 50 ranks i.e., the total count of features chosen. Then the top 3 products which the user might prefer will be recommended.