Prediksi Mood Berdasarkan Pola Aktivitas Fisik pada Remaja Menggunakan Algoritma Machine Learning

Abstract

Penelitian ini bertujuan untuk memprediksi mood remaja berdasarkan pola aktivitas fisik menggunakan algoritma machine learning. Aktivitas fisik sering kali berkorelasi erat dengan kondisi emosional, sehingga pendekatan ini dapat memberikan wawasan baru dalam mendukung kesehatan mental. Data dikumpulkan dari 50 remaja mahasiswa melalui perangkat wearable selama 90 hari, mencakup parameter seperti jumlah langkah, durasi aktivitas, intensitas, dan pola tidur. Sementara untuk data mood diperoleh dari skala BRUMs yang sudah diadaptasi dan divalidasi. Algoritma Decision Tree dan Support Vector Machine (SVM) digunakan untuk membangun model prediksi, dengan evaluasi kinerja berdasarkan metrik akurasi, presisi, recall, dan F1-score. Hasil penelitian menunjukkan bahwa Decision Tree lebih unggul dibandingkan SVM, dengan akurasi sebesar 98,44% dibandingkan 97,45%. Decision Tree juga menunjukkan keunggulan dalam interpretasi model dan efisiensi komputasi, yang penting untuk implementasi aplikasi prediktif real-time. Penelitian ini menyimpulkan bahwa algoritma Decision Tree merupakan pendekatan yang lebih efektif untuk prediksi mood berbasis pola aktivitas fisik pada remaja. Temuan ini diharapkan dapat menjadi dasar pengembangan sistem pendukung kesehatan mental berbasis teknologi wearable.  This study aims to predict adolescents' mood based on physical activity patterns using machine learning algorithms. Physical activity is often closely correlated with emotional states, making this approach potentially valuable in providing new insights to support mental health. Data were collected from 50 university students using wearable devices over 90 days, including parameters such as step count, activity duration, intensity, and sleep patterns. Mood data were obtained using the BRUMs scale, which had been adapted and validated. Decision Tree and Support Vector Machine (SVM) algorithms were employed to develop the predictive model, with performance evaluated based on accuracy, precision, recall, and F1-score metrics. The results showed that the Decision Tree outperformed SVM, achieving an accuracy of 98.44% compared to 97.45%. Decision Tree also demonstrated advantages in model interpretability and computational efficiency, which are crucial for implementing real-time predictive applications. This study concludes that the Decision Tree algorithm is a more effective approach for mood prediction based on physical activity patterns in adolescents. These findings are expected to form the foundation for developing mental health support systems based on wearable technology.