@article{ART003332361},
author={Seungwoo Lee},
title={Intelligent Hot/Cold Data Classification Technique Based on Periodic Epoch Analysis and Logistic Regression for Enhanced Workload Adaptability},
journal={Journal of Internet of Things and Convergence},
issn={2466-0078},
year={2026},
volume={12},
number={2},
pages={49-56}
TY - JOUR
AU - Seungwoo Lee
TI - Intelligent Hot/Cold Data Classification Technique Based on Periodic Epoch Analysis and Logistic Regression for Enhanced Workload Adaptability
JO - Journal of Internet of Things and Convergence
PY - 2026
VL - 12
IS - 2
PB - The Korea Internet of Things Society
SP - 49
EP - 56
SN - 2466-0078
AB - This paper proposes a logistic regression-based, periodic self-adaptive Hot/Cold data classification technique to mitigate the Write Amplification problem inherent in the physical constraints of NAND flash memory. Due to the inability to perform in-place updates and the asymmetry of operational units, NAND flash suffers from significant garbage collection overhead, which can be optimized through accurate data classification. While conventional heuristics suffer from low accuracy in dynamic workloads, deep learning-based approaches impose excessive computational burdens on resource-constrained embedded environments. To address these limitations, we designed an epoch-based periodic analysis model that minimizes real-time inference overhead. The proposed mechanism utilizes a lightweight logistic regression model incorporating frequency, recency, and sequentiality as key features. Furthermore, a self-adaptive mechanism based on oracle labeling—which leverages the access results of the subsequent cycle as ground truth—is introduced to maintain high classification precision even during abrupt workload transitions. Experimental results using MSR Cambridge traces demonstrate that the proposed method improves ROC-AUC by 18% and F1-Score by 12.4% compared to traditional heuristic methods. Most notably, the proposed scheme reduces WAF by up to 40.0% and CPU utilization by more than 77.1%, proving its practical feasibility and efficiency for high-performance embedded storage systems.
KW - NAND Flash Memory;FTL;Hot/Cold Data Classification;Logistic Regression;Self-Adaptive Learning;Write Amplification Factor
DO -
UR -
ER -
Seungwoo Lee. (2026). Intelligent Hot/Cold Data Classification Technique Based on Periodic Epoch Analysis and Logistic Regression for Enhanced Workload Adaptability. Journal of Internet of Things and Convergence, 12(2), 49-56.
Seungwoo Lee. 2026, "Intelligent Hot/Cold Data Classification Technique Based on Periodic Epoch Analysis and Logistic Regression for Enhanced Workload Adaptability", Journal of Internet of Things and Convergence, vol.12, no.2 pp.49-56.
Seungwoo Lee "Intelligent Hot/Cold Data Classification Technique Based on Periodic Epoch Analysis and Logistic Regression for Enhanced Workload Adaptability" Journal of Internet of Things and Convergence 12.2 pp.49-56 (2026) : 49.
Seungwoo Lee. Intelligent Hot/Cold Data Classification Technique Based on Periodic Epoch Analysis and Logistic Regression for Enhanced Workload Adaptability. 2026; 12(2), 49-56.
Seungwoo Lee. "Intelligent Hot/Cold Data Classification Technique Based on Periodic Epoch Analysis and Logistic Regression for Enhanced Workload Adaptability" Journal of Internet of Things and Convergence 12, no.2 (2026) : 49-56.
Seungwoo Lee. Intelligent Hot/Cold Data Classification Technique Based on Periodic Epoch Analysis and Logistic Regression for Enhanced Workload Adaptability. Journal of Internet of Things and Convergence, 12(2), 49-56.
Seungwoo Lee. Intelligent Hot/Cold Data Classification Technique Based on Periodic Epoch Analysis and Logistic Regression for Enhanced Workload Adaptability. Journal of Internet of Things and Convergence. 2026; 12(2) 49-56.
Seungwoo Lee. Intelligent Hot/Cold Data Classification Technique Based on Periodic Epoch Analysis and Logistic Regression for Enhanced Workload Adaptability. 2026; 12(2), 49-56.
Seungwoo Lee. "Intelligent Hot/Cold Data Classification Technique Based on Periodic Epoch Analysis and Logistic Regression for Enhanced Workload Adaptability" Journal of Internet of Things and Convergence 12, no.2 (2026) : 49-56.
