ANALISIS HUBUNGAN MATA KULIAH KOMPUTASI DASAR DENGAN IPK MAHASISWA TEKNIK INFORMATIKA MENGGUNAKAN SUPPORT VECTOR REGRESSION

Student academic performance serves as a key indicator in higher education assessment. For students in Informatics Engineering, foundational computing skills are critical to their academic progression, and these are primarily acquired through first-semester courses. This study proposes a predictive model for Cumulative Grade Point Average (GPA) using the Support Vector Regression (SVR) method with a Radial Basis Function (RBF) kernel. The courses "Introduction to Algorithms," "Computation I," "Computation II," and "Data Structures" were selected as independent variables, as they provide essential computing foundations for subsequent coursework. The dataset comprised 270 records, each containing grades from the aforementioned courses and the corresponding GPA achieved by students in their fourth semester. To ensure data quality, outlier detection was performed using the Z-score method, resulting in a refined dataset of 200 entries. This dataset was then split into 75% for training and 25% for testing. A grid search optimization identified the best hyperparameter combination: C = 100, γ = 0.05, and ε = 0.05. Model evaluation yielded promising results, with a Mean Absolute Error (MAE) of 0.0742, a Mean Absolute Percentage Error (MAPE) of 2.19%, a Mean Squared Error (MSE) of 0.012, and an R² score of 0.8695—indicating strong predictive accuracy. Furthermore, the F-test produced a value of 74.9440, which exceeds the critical F-value of 2.5787, confirming the statistical significance of the independent variables in predicting GPA. This model has the potential to support academic monitoring and enhancement efforts by delivering actionable predictions and insights for the Informatics Engineering program.