Predictive Model for Finance Industry using Machine Learning Algorithms and a Study on Its Performance Analysis

Machine learning which is a process of learning the machine with available data to think like humans is a part of artificial intelligence and has a vast scope as the future technology. It can be used for almost all sectors for predictive analysis, cognitive services, virtual assistance, video surveillance, risk management, fraud detection etc. In this paper, it is concentrating on how the machine learning can be applicable for predicting risk assessment and fraud detection in finance sector which have some imbalanced data sets. Finance sector is the wealth backbone of any country, so risk assessment and fraud detection has great importance. Risk assessment is the process of identifying vulnerabilities to an organization by identifying risk involved in each and every new plans, policies or investments. This paper concentrates on risk level detection of loan application and insurance claim. Fraud detection is the process of finding fraud activities, events or transactions. It can be associated with several fields like money transaction, network intrusion etc. In this paper, fraud money transaction is detected. These all areas contain imbalanced data and these data sets bear with the minority and majority class problems which will hardly affect the prediction accuracy. This paper suggests a predictive model for risk assessment and fraud detection using three efficient machine learning algorithms after applying undersampling technique on data and compares the accuracy difference of them, on imbalanced and resampled data sets with the leading machine learning algorithms Random Forest, SVM (support vector machine) and ANN (Artificial Neural Network). This paper is implemented using Qt IDE and OpenCV library using C++ language. The paper analyses the parameters of each algorithm and tries to find out the solution to increase the efficiency of each algorithm by changing these parameters.