PERBANDINGAN ARSITEKTUR ARTIFICIAL NEURAL NETWORK UNTUK KLASIFIKASI TINGKAT DEPRESI MAHASISWA MENGGUNAKAN DATASET DASS
Abstrak
Student mental health has become an important concern in academia, especially in early detection of depression, anxiety, and stress levels. This study compares the performance of three Artificial Neural Network (ANN) architectures: Multilayer Perceptron (MLP), Radial Basis Function (RBF), and Recurrent Neural Network (RNN) for classifying student depression levels using the Depression Anxiety Stress Scales (DASS) dataset. The dataset consists of 1,812 respondents with 33 features including demographic data and 21 DASS questions. The Knowledge Discovery in Databases (KDD) methodology was applied with stages of data selection, preprocessing, transformation, and data mining. Model evaluation using 5-Fold Cross-Validation with dropout optimization, L2 regularization, and early stopping. The results show that the optimized models achieve better performance than the baseline, with accuracy, precision, recall, and F1-score as evaluation parameters. This research contributes to the selection of appropriate ANN architecture for student mental health detection systems.
Identifier
- DOI: https://doi.org/10.23960/jitet.v14i2.9199
- OpenAlex: https://openalex.org/W7154023748