부산대학교 도서관

This research seeks to compare the accuracy of the novel Random Forest (RF) and Logistic Regression (LR) methods for forecasting software problems which avoids security threats based on a variety of software parameters. The research employed a dataset taken from the Kaggle network, which comprised several software measures such as script lines, program churn, and code intricacy. This study included two machine learning techniques, Logistic Regression and the innovative Random Forest. In the first group, 10 samples were trained employing Logistic Regression, while in the second group, 10 samples were trained employing the novel Random Forest approach. The Python scikit-learn package was utilized for the development of the Logistic Regression and Random Forest methodologies. The evaluation of model accuracy was conducted by employing the testing data subsequent to the training of said models utilizing the training data. Based on a statistical power (G-power) of 80%, a significance level (alpha) of 0.05, and a desired level of type II error (beta) of 0.2, a sample size of 10 per group was judged to be appropriate. The research findings indicate that the Random Forest strategy had a greater accuracy rate of 78.59% in comparison to the Logistic Regression technique, which attained a success rate of 76.54%. In addition, a 95% confidence interval was determined by calculating to ensure that the results were statistically significant values with 0.000 (p