Predicting Future Ratings of Amazon Products Based on the Users' Prior Reviews

The innovative approach, customer-centric service, and diverse product range have established Amazon as the leading e-commerce platform. Every day, thousands of users rely on Amazon.com for online shopping, often depending on user feedback to make purchase decisions. However, this reliance poses a challenge for newly listed products, which typically lack substantial user reviews. This scarcity of feedback can adversely affect buyers' decisions. Our analysis focuses on the large dataset of Amazon Fine Food reviews, an extensive collection of usergenerated reviews on food products sold on the platform. The primary objective is to develop a method for estimating ratings for products with few reviews. We aim to create a system that provides valuable insights, even with limited data. Despite the scarcity of information, our model has successfully generalized unseen data, demonstrating low prediction errors. Across the 19 datasets analyzed, the Mean Square Error, normalized for the total number of datasets, was a low 0.153. This shows our model's effectiveness in offering helpful information and predictions under data constraints. In the future, we want to improve the model by fine-tuning the hyperparameters.