Use of collaborative filtering algorithms to improve the e-commerce performance

Idea: To analyze the importance of the recommendation systems algorithms in the scope of increasing the e-commerce performance, addressing the customer's experience enhancement and sales increase. To discover what are the best algorithms to be used in an e-commerce website for precise results. Data: The Movie Lens dataset, collected from September 19, 1997 to April 22, 1998. The dataset contains 100,000 ratings (1-5) from 943 users on 1664 movies. Tools: The software used to apply the algorithms on the dataset is R, the statistical language that offers support in building recommendation systems, by making the simulations to be real. What's new: The article analyzes the three of the most used algorithms on e-commerce websites (random, popular, user-based collaborative filtering), applied on a real dataset (MovieLens). The article, thus, builds a comparison between these three algorithms in the scope of enhancing the prediction accuracy for customers. Furthermore, it is provided a description of the benefits of using recommendation systems in the e-commerce. So What: These findings add value for the e-business owners by explaining the importance of using recommendation systems and by providing an analysis of the algorithms that offer the most accurate predictions. Contribution: It is discovered the algorithm that offers the most accurate suggestions, based on an analysis using two methodologies: the ROC Curve and the precision and recall balance. The article results provide a rank of the algorithms to be used in any e-commerce website that wants to enhance customer experience and increase sales.