Our Project

This project leverages supervised machine learning, specifically k-nearest neighbors, to identify potential fraudulent activities within bank payments. The dataset comprises transactional information such as customer demographics, merchant details, transaction amounts, and a fraud indicator.

Through an exploratory data analysis (EDA), insights into transaction patterns, customer behavior, and categorical distributions are uncovered. Using multiple types of models including logistic regression, decision trees, and random forests but focusing on k-nearest neighbors, the project aims to segment transactions into distinct groups based on transaction attributes, seeking anomalous clusters that deviate significantly from regular transaction patterns.

Our Goal

The project's ultimate goal is to upgrade the existing frameworks used for identifying potential fraudulent transactions, offering financial institutions a proactive approach to mitigate risks associated with fraud.