Identifying Corporate Financial Fraud using Pearson Universal Kernel function-based Support Vector Machine

Corporate Financial fraud has significantly affected the endurable development of the financial industries as the critical issue globally. However, it is slightly difficult to determine the frauds with significantly imbalanced datasets due to the rate of non-fraud industries which is too maximum as compared to the fraudulent ones. An intelligence financial statement fraud detection system has significantly developed to help the decision-making for the stakeholders. Therefore, this research proposes the Pearson Universal Kernel function-based Support Vector Machine (PUK-SVM) approach for Identifying corporate financial fraud. A PUK kernel is highly significant at handling nonlinear relationships in data, designing it well-appropriated for the challenging financial data where fraud patterns. The pre-processing step like Min-max normalization approach is performed for normalize the input data from the collected financial data. The Word-to-Vector (Word2Vec) approach is utilized for the feature extraction process from the pre-processed data. Finally, the proposed PUK-SVM approach is utilized to perform into binary classification such as fraud and non-fraud. The experimental outcomes demonstrates that the proposed PUK-SVM method attains the better accuracy result of 96.39% respectively as compared to the existing approaches like SVM and K Nearest Neighbour (KNN).