Exploring Collaborative Filtering Methods for Product Recommendations in E-Commerce: A Study Using Amazon and BigBasket Datasets

Recommendation systems play a significant role in today’s age of technology. These systems use machine learning algorithms and act as an information filtering tool to help users discover relevant and personalized content, products, and services based on their past behaviours and interests. Applications of recommendation systems range from e-commerce, news and article recommendation, social media, healthcare and education. Online shopping has rapidly expanded, which has led to an exponential rise in the number of items and options available. Since the e-commerce sites generate huge amount of data users may find it difficult in selecting the right product for their needs. To address this issue e-commerce sites, need an effective and accurate recommendation techniques. Many techniques have been proposed for building recommender systems and its effectiveness varies depending on the application domain and the dataset used. This paper compares the performance analysis of different collaborative filtering algorithms which includes SVD, SVD++, Slopeone, KNN, KNN with Means and KNN with Zscore using two Amazon and Big Basket datasets. According to experimental studies SVD++ out performed the other algorithms with a reduced error rate. In this work we have used MAE and RMSE as evaluation metrics to analyse the algorithms performance.