The article discusses fraud detection and how to fully automate it. It is now crucial for fraud detection in every bank. There is a considerable increase in fraud, which causes the banks significant losses. Transactions present particular difficulties for fraud exposure because there is no short-term processing available. A feasibility analysis of the selected fraud detection systems is the first task. These transactions are to be tested individually and continued with the aid of models. We first establish a detection task, including the dataset's characteristics, the chosen metric, and any controls for such unbalanced datasets. This leads to the discovery that the dataset's underlying pattern produces the following results:For instance, cardholders may alter their buying patterns over time, while fraudsters may alter their strategies. Later, we presented a number of techniques for obtaining credit card sequential features. Financial fraud is the practise of acquiring financial benefits by dishonest and illegal ways. Financial fraud has recently become a severe concern to businesses and organisations, which is defined as the employment of dishonest means to earn financial profits. Financial fraud continues to have a detrimental influence on society and the economy despite numerous efforts to stop it because the daily losses from fraud amount to substantial sums of money. Many years ago, the first fraud detection techniques were established. The bulk of outdated processes still use manual labour, which is not only expensive, imprecise, and time-consuming, but also unworkable. More studies are being conducted, however they have no effect in lowering fraud-related losses. This study uses the Random Forest Classifier Machine Learning Algorithm to provide a novel model of fraud detection on bank payments. We have shown that our proposed system, which uses the Banksim dataset, is superior to the existing one by reaching train and test accuracy of 99%.