@article{ART003280460},
author={Maengsu Kim and Hyun Sim},
title={Missing/Noise Smart Farm Time Series Sensor Data A Study on a Lightweight Deep Learning Model for Restoration},
journal={Journal of The Korea Society of Computer and Information},
issn={1598-849X},
year={2025},
volume={30},
number={12},
pages={147-157}
TY - JOUR
AU - Maengsu Kim
AU - Hyun Sim
TI - Missing/Noise Smart Farm Time Series Sensor Data A Study on a Lightweight Deep Learning Model for Restoration
JO - Journal of The Korea Society of Computer and Information
PY - 2025
VL - 30
IS - 12
PB - The Korean Society Of Computer And Information
SP - 147
EP - 157
SN - 1598-849X
AB - This study compares traditional interpolation methods (Linear, Spline, KNN, Kalman) and lightweight deep learning models (LSTM-AE, Masked-AE, TCN-Lite) under identical conditions, and verifies real-time applicability on a Raspberry Pi 5 edge device via ONNX inference. For 1-minute-interval multivariate sensors (temperature, humidity, CO₂, soil moisture, EC, solar irradiance, irrigation, yield), we constructed a benchmark by injecting 5% point-wise missingness, 2% six-minute block missingness, Gaussian noise (σ = 0.05), and drift (±5%). Consequently, the deep-learning group outperformed the traditional methods with reduced RMSE/MAE and improved R² and SNR/PSNR, with TCN-Lite achieving the best performance at approximately R² ≈ 0.874. In edge experiments, both LSTM-AE and TCN-Lite showed < 2% performance loss compared to PC runs, demonstrating feasibility for real-time reconstruction. This study presents (i) a single pipeline that spans traditional methods–lightweight deep learning–edge deployment, (ii) a multi-dimensional metric framework that evaluates accuracy and efficiency together, and (iii) design guidelines for practical field deployment.
KW - Time-Series Restoration;ONNX Inference;Edge Computing;Mode Lightweighting
DO -
UR -
ER -
Maengsu Kim and Hyun Sim. (2025). Missing/Noise Smart Farm Time Series Sensor Data A Study on a Lightweight Deep Learning Model for Restoration. Journal of The Korea Society of Computer and Information, 30(12), 147-157.
Maengsu Kim and Hyun Sim. 2025, "Missing/Noise Smart Farm Time Series Sensor Data A Study on a Lightweight Deep Learning Model for Restoration", Journal of The Korea Society of Computer and Information, vol.30, no.12 pp.147-157.
Maengsu Kim, Hyun Sim "Missing/Noise Smart Farm Time Series Sensor Data A Study on a Lightweight Deep Learning Model for Restoration" Journal of The Korea Society of Computer and Information 30.12 pp.147-157 (2025) : 147.
Maengsu Kim, Hyun Sim. Missing/Noise Smart Farm Time Series Sensor Data A Study on a Lightweight Deep Learning Model for Restoration. 2025; 30(12), 147-157.
Maengsu Kim and Hyun Sim. "Missing/Noise Smart Farm Time Series Sensor Data A Study on a Lightweight Deep Learning Model for Restoration" Journal of The Korea Society of Computer and Information 30, no.12 (2025) : 147-157.
Maengsu Kim; Hyun Sim. Missing/Noise Smart Farm Time Series Sensor Data A Study on a Lightweight Deep Learning Model for Restoration. Journal of The Korea Society of Computer and Information, 30(12), 147-157.
Maengsu Kim; Hyun Sim. Missing/Noise Smart Farm Time Series Sensor Data A Study on a Lightweight Deep Learning Model for Restoration. Journal of The Korea Society of Computer and Information. 2025; 30(12) 147-157.
Maengsu Kim, Hyun Sim. Missing/Noise Smart Farm Time Series Sensor Data A Study on a Lightweight Deep Learning Model for Restoration. 2025; 30(12), 147-157.
Maengsu Kim and Hyun Sim. "Missing/Noise Smart Farm Time Series Sensor Data A Study on a Lightweight Deep Learning Model for Restoration" Journal of The Korea Society of Computer and Information 30, no.12 (2025) : 147-157.
