The banking industry is an important part of modern actions as it manages the movement of funds between different parties. However, this area is synonymous with some cases of fraud where people are being swindled their money, illegal transactions are being made and others.

The complexity of ensuring that transactions stay legitimate has since made it almost impossible to regulate fraud in this industry correctly. This report presents an approach that utilizes Machine Learning techniques to build a model that detects fraudulent transactions and flags them. The approach utilizes a dataset that contains a collection of observation points on transactions and which can be useful in understanding the nature of transactions. The data is clearly imbalanced, an issue fixed in the data preparation section of the analysis. The models of choice are CatBoost classifier, Decision Trees classifier and Random forest classifier. The CatBoost classifier performs best, followed by the decision tree and then the Random forest. However, they all present quite high accuracy rates of classification at over 90%. The high accuracy results of the models are indicative of their readiness to use in a real-world setting. This performance means that the likelihood of a fraudulent case passing through is quite low.

The recommendation to utilize the approach is to add more variables for better descriptions of fraud transactions and improve the results. It is also possible to improve the results by increasing the size of the dataset or using more models and comparing the results. A combination of models can also turn out to be a good approach for much better results for use in the real world.

Publication Date

Fall 2022

Document Type

Master's Project

Student Type


Degree Name

Professional Studies (MS)

Department, Program, or Center

Graduate Programs & Research (Dubai)


Sanjay Modak

Advisor/Committee Member

Ehsan Warriach


RIT Dubai