Optimizing Customer Experience in Online Shopping: A Hyperparameter Tuning Approach

After buying a product, the majority of buyers don't leave reviews, ratings, or comments, which makes it difficult for any e-commerce site to function. If a business can forecast whether a customer will like or dislike a product, it can use that information to suggest other comparable products and make decisions about whether or not to sell a certain product. This study's main goal is to forecast customer happiness with a product, which may be done by projecting the review rating that a client will provide after making a purchase. Using our dataset, we evaluate each classifier's performance using hyper parameter tuning procedure. Grid Search CV method is used with five machine learning algorithms— Naive Bayes , Logistic Regression, Decision Tree, Random Forest and Gradient Boost Decision Tree (GBDT) to find the customer satisfaction . The results showed that, GBDT outperforms the rest of the model with highest accuracy of 86.57. While both random forest and logistic regression perform reasonably well, but GBDT is more resilient to outliers.