EVALUASI EFEKTIVITAS SISTEM DETEKSI PENIPUAN BERBASIS AI MENGGUNAKAN METODE REGRESI LOGISTIK UNTUK MENINGKATKAN KEAMANAN TRANSAKSI PADA STARTUP FINANCE

Abstract

This research evaluates the effectiveness of an AI-based fraud detection system in a finance startup using logistic regression methods. The primary objective is to assess whether this model is feasible for enhancing transaction security. The research methodology employed is quantitative, utilizing historical transaction data. The test results indicate that the logistic regression model can accurately predict the company's financial conditions, with an accuracy of 95%. However, there are challenges in detecting fraudulent transactions, as evidenced by the low recall for fraudulent transactions. The model evaluation was conducted using several performance metrics such as accuracy, precision, recall, and F1-score, as well as model fit tests including Hosmer and Lemeshow’s Goodness of Fit Test, Omnibus Test of Model Coefficients, and -2 log-likelihood. In conclusion, this model has the potential to enhance transaction security, but it still requires improvements to more effectively detect fraudulent transactions