Confirmed quality aware recommendations using collaborative filtering and review analysis

Abstract

: Recommendation Systems (RS) save the time of users in their hectic life schedules for purchasing their interested products.RS faces challenges of data sparsity, cold start, efficiency of prediction of products and hence the proposed system is making use of Multi-Kernel Fuzzy C Means (MKFCM) clustering to group together similar users having similar age, occupation and gender into clusters. Clusters of similar users are optimised using the Fruit Fly (FF) optimisation algorithm which gives high cluster accuracy and dynamically created sub-clusters of similar users with their favourite products, overcome sparsity issue which make the analysis easy. Collaborative Filtering (CF), one of the filtering method of RS is used to predict products for target users. This RS gains user’s faith by additionally performing analysis of textual reviews using optimised Artificial Neural Network (ANN) to recommend the highest quality products, thus dual tested and quality confirmed products are recommended to the user. Experimentation is done on a standard Movilense data set used by many researchers to prove the efficiency of this RS and reviews of all users are extracted from online search engines for product quality analysis before recommendation. Experimentation proves higher recall and accuracy than existing recommendation systems.