Sentiment Analysis-Based Hybrid Collaborative Filtering Recommendation Algorithm for Consumer Decision

Abstract

A recommendation system can suggest things for diverse user interests based on several sources of information. Most recommendation systems rely heavily on the collaborative filtering (CF) approach, in which the user's preference data is combined with that of other users to make predictions about further things the consumer can be interested in. In this study, a novel weighted recommendation system is created for better consumer decisions using CF.  An equation to calculate the weight of both the product and review and an equation to calculate the similarity between consumer’s review is created in the methodology. Random Forest (RF), Support Vector Machine (SVM), and Logistic Regression (LR) are used in the methodology to ensemble the model. Ensemble Classifiers (RF+SVM+LR) are taken into consideration to implement the results of the methodology to get better results in comparison to the previous studies. The proposed model is trained and tested using an open-source dataset that is available on the website of Kaggle. Numerical analysis of the proposed model shows that it performed better than other conventional methods in terms of accuracy (0.821), precision (0.802), recall (0.821), F-measure (0.833), and error rate (0.100), and many more.