Assessing the Impact of Traffic-Related Challenges on Academic Performance: A Comparative Study of Statistical and Machine Learning Models

This research evaluates the effectiveness of statistical and machine learning models in understanding the impact of traffic-related issues on academic performance. We analyzed a dataset of 80 survey responses exploring relationships between traffic frequency, stress levels, transportation costs, and academic outcomes. Our analysis included Ordinary Least Squares (OLS) regression, Poisson regression, Bayesian Ridge regression, and machine learning techniques like Support Vector Regression, Ridge Regression, Random Forest, and XGBoost. Data preprocessing involved MinMax scaling for numerical features and encoding for categorical variables. We focused on metrics such as missed classes and exam-related stress, assessing model performance using R², Mean Squared Error (MSE), and Mean Absolute Error (MAE). The OLS regression model excelled, achieving an R² of 0.743, MSE of 0.0192, and MAE of 0.1123. Both Bayesian Ridge and Poisson regression demonstrated solid performance. This study underscores the critical impact of traffic-related factors on academic success and illustrates the varying effectiveness of different predictive models. While machine learning models showed potential, they could not perform as well as statistical models due to data limitations. Therefore, additional tuning is necessary to optimize their performance.