An Ensemble Model-Based Recommendation Approach for Consumer Decision-Making System

Abstract

A recommendation system can suggest items aligned with diverse user interests by leveraging multiple sources of information. While many recommendation systems heavily rely on the collaborative filtering (CF) approach—where user preference data is combined with others to predict additional items of potential interest—this study introduces a novel weighted recommendation system to enhance consumer decision-making using CF. The methodology includes the development of equations to calculate the weights for both the product and review, as well as to determine the similarity between consumer reviews. To ensemble the model, Random Forest (RF), Support Vector Machine (SVM), and Logistic Regression (LR) are employed in the methodology. The study considers Ensemble Classifiers (RF+SVM+LR) to implement the results, aiming for improved outcomes compared to prior research. The proposed model is trained and tested using an open-source dataset on Kaggle's website. Numerical analysis of the proposed model reveals superior performance, outperforming conventional methods in terms of accuracy (0.821), precision (0.802), recall (0.821), F-measure (0.833), error rate (0.100), and more.