Machine learning approach on apache spark for credit card fraud detection

Abstract

In the digital era, industries such as banking and financial organizations are facing different challenges with transaction-related activities. One of the significant challenges in financial organizations is Credit card fraud. In order to identify credit card fraud activities. In this paper, we employ an integrated hybrid approach using Apache Spark. The proposed hybrid approach is the integration of K-Means and C5.0 decision tree with an adaptive method, which is examined through Hadoop and Spark. Using K- Means, we find the closest clusters, and with the rules of a decision tree, each normal and fraud instance in the dataset is classified. This model is evaluated on one million synthetic datasets and achieved a good classification rate. We present our model with detailed experimental results and comparison with other models. This model is suitable for computationally complex datasets, and it can be applied to various fields for anomaly detection on big data.