Predicting Fraudulence Transaction under Data Imbalance using Neural Network (Deep Learning)

The number of financial transactions has the potential to cause many violations law (fraud). Conventional machine learning been widely used, including logistic regression, random forest, and gradient boosted. However, can work as long dataset contains fraud. Many new technology companies need anticipate for fraud, which they have not experienced much. This a crime also be by old service providers with low frequency previous With data imbalance, traditional learningis likely produce false negatives so that do accurately predict study optimizes approach based on Neural Networks improve model accuracy through integration KNIME Python Programming KERAS TensorFlow models. conducts comparative analysis scrutinize performance Adam Adamax Optimizer. Using from European cardholders in 2013, this proves workflows neural network algorithms detect up 95% even very small fraud sample only 0.17% or 492 284,807 transactions. In addition, optimizer performs higher than optimizer. implication is supervisory innovation developed minimize transaction crimes services sector.