@article{ART003317602},
author={Yong-Je Ko and Ho-Young Kwak},
title={Time-Period-Aware Difficulty Assessment and Modeling Framework for Improved Building Energy Consumption Prediction},
journal={Journal of The Korea Society of Computer and Information},
issn={1598-849X},
year={2026},
volume={31},
number={3},
pages={89-96}
TY - JOUR
AU - Yong-Je Ko
AU - Ho-Young Kwak
TI - Time-Period-Aware Difficulty Assessment and Modeling Framework for Improved Building Energy Consumption Prediction
JO - Journal of The Korea Society of Computer and Information
PY - 2026
VL - 31
IS - 3
PB - The Korean Society Of Computer And Information
SP - 89
EP - 96
SN - 1598-849X
AB - Building energy consumption prediction is essential for energy management and optimization, but existing research applies a single model across all time periods, failing to adequately reflect the different consumption patterns and prediction difficulties of each period. This study develops a new metric called the Prediction Difficulty Index to quantify time-period-specific prediction difficulty and empirically demonstrates the necessity and effectiveness of differentiated modeling strategies. Our analysis quantified that the PDI-based prediction difficulty of Transition periods is 4.36 times higher than Off-peak periods(MAPE-based:4.47 times) using the Prediction Difficulty Index, which integrates variability and prediction error. Statistical tests showed highly significant differences between time periods (ANOVA: F=26.35, p<0.000001; Cohen's d=1.95), and SHAP analysis confirmed that different features play important roles in prediction for each time period. We developed period-specific models and evaluated them using Leave-One-Out Cross-Validation (LOOCV), achieving a 30.58% improvement in MAPE compared to the baseline LightGBM model. This demonstrates that even simple models (Linear Regression, Ridge Regression) can outperform complex single models when tailored to time-period characteristics, even with small data (94 samples).
KW - Energy Consumption Prediction;Time-Period Modeling;Prediction Difficulty Index;LightGBM;SHAP Analysis;Differentiated Modeling
DO -
UR -
ER -
Yong-Je Ko and Ho-Young Kwak. (2026). Time-Period-Aware Difficulty Assessment and Modeling Framework for Improved Building Energy Consumption Prediction. Journal of The Korea Society of Computer and Information, 31(3), 89-96.
Yong-Je Ko and Ho-Young Kwak. 2026, "Time-Period-Aware Difficulty Assessment and Modeling Framework for Improved Building Energy Consumption Prediction", Journal of The Korea Society of Computer and Information, vol.31, no.3 pp.89-96.
Yong-Je Ko, Ho-Young Kwak "Time-Period-Aware Difficulty Assessment and Modeling Framework for Improved Building Energy Consumption Prediction" Journal of The Korea Society of Computer and Information 31.3 pp.89-96 (2026) : 89.
Yong-Je Ko, Ho-Young Kwak. Time-Period-Aware Difficulty Assessment and Modeling Framework for Improved Building Energy Consumption Prediction. 2026; 31(3), 89-96.
Yong-Je Ko and Ho-Young Kwak. "Time-Period-Aware Difficulty Assessment and Modeling Framework for Improved Building Energy Consumption Prediction" Journal of The Korea Society of Computer and Information 31, no.3 (2026) : 89-96.
Yong-Je Ko; Ho-Young Kwak. Time-Period-Aware Difficulty Assessment and Modeling Framework for Improved Building Energy Consumption Prediction. Journal of The Korea Society of Computer and Information, 31(3), 89-96.
Yong-Je Ko; Ho-Young Kwak. Time-Period-Aware Difficulty Assessment and Modeling Framework for Improved Building Energy Consumption Prediction. Journal of The Korea Society of Computer and Information. 2026; 31(3) 89-96.
Yong-Je Ko, Ho-Young Kwak. Time-Period-Aware Difficulty Assessment and Modeling Framework for Improved Building Energy Consumption Prediction. 2026; 31(3), 89-96.
Yong-Je Ko and Ho-Young Kwak. "Time-Period-Aware Difficulty Assessment and Modeling Framework for Improved Building Energy Consumption Prediction" Journal of The Korea Society of Computer and Information 31, no.3 (2026) : 89-96.
