Nowadays, money transfer through the internet has become so popular because of its convenience and speed which makes users’ lives easier. Even so, the safety of these transactions has been threatened by illegal activities causing great difficulty and loss for users. One of those unauthorized actions is fraud through credit cards used for financial transactions on online platforms. Therefore, research in detecting and early warning of fraudulent transactions through credit cards is essential today. In this paper, we propose a new approach for the task of early detection of fraudulent transactions based on a combination of two main methods, behavioral analysis techniques and supervised machine learning algorithms. Specifically, based on the behavioral analysis technique proposed in this paper, we have selected and extracted new features. These are features that have not been reported in previous studies. In addition, for the classification method, we propose to use a new advanced supervised machine learning algorithm, XGBoost. This is a newly researched and proposed machine learning algorithm. Based on the proposed approach in this paper, we have not only succeeded in synthesizing, analyzing and extracting the anomalous behavior of fraudulent transactions but also improved the efficiency of detecting suspicious transactions. Some experimental scenarios proposed in the paper have proven that our proposal in this paper is not only meaningful in terms of science but also in practical terms when the results of the paper have been proven more effective than some other approaches