Improved personalized recommendation system with better user experience
Pratik Ghanwat, Anu Mary Chacko · 2017
With the development of Internet e-commerce, shopping online has become very popular. Online stores have the advantage of having more items in their catalogue without worrying about shop constraints. In this scenario, recommendation systems are useful as they can help users in identifying items that they might be interested in. After shopping users usually provide feedback in the form review and ratings. This feedback is very useful for recommendation systems. Most of the current implementations of recommendation engines makes use of ratings only. The review text contain important aspect on why a particular rating was given. If the review text can be analyzed and used for recommendation, the user experience can be improved. In this paper we propose an approach to identify `aspects' in the review and build user and item profiles to reflect `aspects'. The users' preference to `aspects' are considered during recommendation. The proposed approach was tested with data from Amazon dataset and our proposed recommendation system showed RMSE MAE value below one. The salient feature of this approach is that it combines review text and rating to remove sparsity and cold start problem in a limited sense.