An Analysis of Various Algorithmic Behaviors in Detecting a Financial Fraud
T. Sreeja Reddy, Guttula Nookaraju, Kummari Vikas, Sachi Nandan Mohanty, Jagruthi Anagandula, Mohammed Sami Ahmed · 2022 13th International Conference on Computing Communication and Networking Technologies (ICCCNT) · 2022
The most interesting thing about occurrences of fraud is that they tend to have a pattern. Though nonsingular or linear, the patterns can be very divergent and have a wide leap of repetition. In recent decades, there’s been a growing reliance on cloud and mobile computing, and though there’s security, frauds tend to evolve with new patterns. Developing a fast-paced identifier of these patterns tends to help in detecting fraud beforehand with higher accuracy. The project is an analysis of how different algorithms and their combinations behave while detecting fraud. We used an Automobile insurance claim data-set operated under various algorithms like XG-Boost, Linear Discriminant Analysis, Gradient Boosting Classifier, Bernoulli NB, Bagging Classifier, Decision Tree Classifier, Ada-Boost Classifier, Random Forest Algorithm, K Neighbours Classifier, Support Vector Machine, and Logistic Regression Algorithm. The model responds better with XG-Boost followed by Linear Discriminant Analysis. This could be paired up with the idea that gradient-based algorithms work faster and are more accurate and can be integrated with an improvised model sequencing the algorithms that predict fraud with minimum fallback.