Deep Learning of Mainshock and Aftershocks After 2015 Ranau Moderate Earthquake Using Wavelet Method

The 2015 Ranau earthquake sequence, consisting of a Mw 6.0 mainshock followed by more than 80 aftershocks, provides a critical dataset for understanding ground-motion behavior in a shallow active fault system. This study analyses waveform characteristics from the closest seismic station using segmented time-series analysis (20–80%) to evaluate temporal variations in shaking intensity. Acceleration records were selected as the primary input for signal processing, enabling accurate estimation of high-frequency content, seismic energy, and time–frequency evolution using the continuous wavelet transform (CWT). Statistical parameters including mean, standard deviation, RMS, skewness, and kurtosis were computed for each segment to assess amplitude variability and impulsive shaking. Results show that the mainshock exhibits markedly higher RMS, standard deviation, and kurtosis, especially within the 50% segments, indicating concentrated strong motion dominated by S-wave energy. In comparison, the 40% segment of Mw 4.8 and Mw 5.1 aftershocks display low and stable statistical values and moderate shaking with identifiable mid-segment intensification. CWT analysis further confirms stronger frequency content and energy release in the mainshock relative to the aftershocks. These findings highlight the importance of segmented waveform analysis for revealing intra-event shaking evolution and provide improved understanding of ground-motion characteristics relevant to structural response, seismic hazard assessment, and future data-driven ground-motion prediction frameworks.