Intelligent Financial Fraud Detection Using Artificial Bee Colony Optimization Based Recurrent Neural Network

Frauds don't follow any recurring patterns. They require the use of unsupervised learning since their behaviour is continually changing. Fraud-sters have access to the most recent technology, which gives them the ability to defraud people through online transactions. Fraudsters make assumptions about consumers' routine behaviour, and fraud develops swiftly. Unsupervised learning must be used by fraud detection systems to recognize online payments since some fraudsters start out using online channels before moving on to other techniques. Building a deep convolutional neural network model to identify anomalies from conventional competitive swarm optimization pat-terns with a focus on fraud situations that cannot be identified using historical data or supervised learning is the aim of this paper Artificial Bee Colony (ABC). Using real-time data and other datasets that are readily available, the ABC-Recurrent Neural Network (RNN) categorizes fraud behaviour and compares it to the current algorithms. When compared to the current approach, the findings demonstrate that the accuracy is high and the training error is minimal in ABC_RNN. In this paper, we measure the Accuracy, F1 score, Mean Square Error (MSE) and Mean Absolute Error (MAE). Our system achieves 97% accuracy, 92% precision rate and F1 score 97%. Also we compare the simulation results with existing methods.