A large number of people now engage in electronic commerce, thanks to the proliferation of contemporary gadgets and their advanced applications, as well as technological breakthroughs that have enabled products and services to be available over the Internet. Customers may feel overwhelmed and find it challenging to locate the ideal product due to the huge selection and diversity of products available on e-commerce websites. Due to the increased competition among international commercial sites brought about by these reasons, it is more important than ever to function profitably and efficiently. The performance of e-commerce platforms is something that can be enhanced by machine learning-based systems, which make it easier for users to locate products that suit their preferences. Many different machine learning algorithms serve this function. A major purpose of this study work is to assist proprietors of E-commerce websites based on machine learning methods to analyse consumer behaviour. An e-commerce site's performance can be improved with the help of a system developed in this study; this system uses an analysis of customers' behaviours to inform decision-making. To tackle these problems and generate highly accurate predictions with grid search for stacked model results, machine learning-based stacking algorithms (XGBoost and CatBoost) are developed. Utilising criteria like F1-Score, ROC-AUC, Accuracy, Precision, and Recall to assess system performance, experimental findings demonstrate the project's contribution. According to the experimental results, applying ML techniques enhances decision-making, which in turn increases the accuracy of suggestion lists that are provided to customers.

E-Commerce Websites, Customer Behavior, Machine Learning, Artificial intelligence, Supervised machine learning algorithms, Stacking, XGBoost and CatBoost, GridSearchCV,