CharlesChrisstianto

This study explores next-day gold price direction prediction using supervised machine learning. Daily Indonesian gold price data are transformed into features such as log returns, moving-average ratios, and short-term volatility, then classified using Naïve Bayes, Logistic Regression, and Support Vector Machine. Among the evaluated models, Logistic Regression achieved the strongest overall performance, with an F1-score of 0.737 and accuracy of 58.3%, showing the potential of engineered financial features for short-term directional forecasting.

This research presents an embedded system for early sleep apnea screening by combining physiological monitoring with audio-based snoring detection. The system uses MFCC feature extraction and a Convolutional Neural Network (CNN) to classify snoring sounds, while sensors connected to an ESP32 and Raspberry Pi monitor parameters such as SpO₂, heart rate, respiratory activity, and body movement. The optimized CNN achieved 99.38% training accuracy and 98.50% testing accuracy, demonstrating the system’s potential as a portable, non-invasive solution for preliminary sleep apnea monitoring outside clinical environments.

V-AIR (Voice-Activated Intelligent Remote) is a voice-based UAV control system designed to make drone operation more intuitive and accessible. The system uses Mel-Frequency Cepstral Coefficients (MFCC) for speech feature extraction and an SVM with an RBF kernel to recognize six basic flight commands: takeoff, land, forward, backward, left, and right. The developed model achieved 98.66% training accuracy and 85.23% testing accuracy, demonstrating the potential of voice commands as a practical hands-free alternative to conventional UAV controllers.