Variance based product recommendation using clustering and sentiment analysis

With the growth of technology, new products are introduced in the market everyday. This rise in products has initiated competition among different manufacturers. Industries use different strategies to make their products more appealing to the customers and survive in the market. As a result, when launching any new product, it is necessary to analyze the reason behind the success of top competing products in the market having similar properties. This information can be useful for the success of the newly launched product. In view of this challenge, a variance-based product recommendation (VPR) approach is proposed, which aims to find top competitors for any newly launched product having similar description. VPR is based on ratings of a product. Thus, rating by a user is found by combining the user-given rating and the rating predicted from the corresponding review text using lexicon-based sentiment analysis. To implement the approach, the available products are firstly divided into clusters based on description. VPR is then applied on one of the clusters to which the new product belongs to get the top competitors for that product. The approach is verified by conducting tests on a product data set of Amazon reviews obtained from SNAP web data, mashup data from ProgrammableWeb as well as product data from Flipkart.