Integrating Sentiment Analysis in Book Recommender System by using Rating Prediction and DBSCAN Algorithm with Hybrid Filtering Technique

Abstract

The recommender system (RS) shows a personalized recommendation by separating the data based on what clients like. Nowadays, people want to buy the most popular products and services to spend the least time shopping. The products are suggested based on what the customer has bought before, what they like, what they say, their profile, the best feature on a website, etc. In this article, we show a hybrid filtering method for book recommendations. That uses the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering technique to meet each person's needs. In addition, reviews of the books are taken into account to figure out the rating. These are grouped into two groups: reviews with ratings and ratings and reviews without reviews and ratings (missing data). In a complete review, the sentiment score is calculated by adding the text from the study that shows how people feel about it. The feeling could be either good or bad. In an incomplete review, the rating is based on the user's demographic information (age, gender, locality & profession). This article also looks at the different types of similarity measures, such as Adjusted Cosine, Pearson Correlation, Euclidean, Manhattan, and Jaccard Similarity. The proposed method is tested on the Amazon book dataset. The RS error is calculated using Root Mean Square Error (RMSE) and Mean Square Error (MSE). The results show that the suggested method has a lower error rate with RMSE (2.63), MSE, and MSE (3.15). This method solves the problems of a cold start and a lack of data while giving them valuable books and amenities. The accuracy of recommendations is measured by precision, recall, and the F-measure.