Pandemic Covid-19 allowed electronic trading to grow even faster, leading to a large increase in online fraud. This is caused by businesses and customers of great financial danger. To make the digital marketplace safe, we need improved fraud detection algorithms. However, these systems often prevent the fact that real world data is difficult. In this study, a data set of electronic trading was used to assess various “ML algorithms to detect fraud, including logistics regression, decisionmaking tree, random forest, naive gulf, Support Vector Machine (SVM), Artificial Neural Networks (ANN), K-Nearest Neighbours (KNN), and boosting techniques such as CATBoost, AdaBoost, Gradient Boosting, and XGBoost”. To overcome class imbalance, the “Synthetic Minority Oversampling Technique (SMOTE)” was used to solve the problem of the class imbalance. File techniques were also used for more accurate predictions. The voting classifier, “which combines bagging with random forest and increased decision -making tree, provided the best results with 100% accuracy”. The results show that file DL methods are very good to observe fraud and can be used to protect e -trading websites such as eBay and Facebook. Overall, this research shows how important the powerful machine learning methods are safer and more credible for the rapid growing digital marketplace