Machine Learning-based GST Fraud Detection using Support Vector Machine (SVM) for Enhanced Tax Governance

The Goods and Services Tax (GST) was implemented as a significant tax reform to in-crease revenue collection, simplify taxation, and promote transparency. Nonetheless, GST fraud is still a major problem that has an impact on both economic stability and govern-ment revenue. False invoicing, taking advantage of the Input Tax Credit (ITC), and circu-lar trade are examples of fraudulent acts that cause significant financial losses and inter-fere with ethical company procedures. In addition to lowering tax collections, these fraud-ulent techniques make it more difficult for regulatory bodies to conduct thorough audits and investigations. This study suggests a machine learning-based fraud detection solution that employs the Support Vector Machine (SVM) classifier, which achieves an accuracy of 97.51%. Through the examination of important financial indicators, including invoice amounts, taxable values, ITC claims, and buyer-seller transaction patterns, the system can accurately spot suspicious activity. To enhance model performance, data preparation methods such as Recursive Feature Elimination (RFE) for feature selection and SMOTE for class imbalance handling are used. The re-search's conclusions help to strengthen tax governance and guarantee a stronger, more open GST system.