Article Per Year (5 Year)
p-Index From 2021 - 2026
0.23
P-Index
This Author published in this journals
All Journal Knowbase : International Journal of Knowledge in Database
Irfan Fahmi Ahmadi
Telkom University
Computer Science & IT

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

 1 
Lightweight Hybrid SVM-LSTM Edge AI for Real-Time Smart Greenhouse Anomaly Detection Irfan Fahmi Ahmadi; Seno Adi Putra; Atam Rifa'i Sujiwanto
Knowbase : International Journal of Knowledge in Database Vol. 6 No. 1 (2026): June 2026
Publisher : Universitas Islam Negeri Sjech M. Djamil Djambek Bukittinggi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30983/knowbase.v6i1.11389

Abstract

Smart greenhouse systems require an artificial intelligence model that can transform sensor records into real-time abnormal-condition decisions while remaining practical for edge computing devices. This paper proposes a lightweight Hybrid SVM-LSTM Edge AI model for real-time smart greenhouse anomaly detection. The model integrates packet validation, deterministic feature fusion, SVM-based discriminative scoring, and LSTM temporal sequence learning to classify greenhouse states as normal or abnormal. The feature representation combines current sensor values, short-term sensor changes, sampling-gap information, and daily periodic patterns so that the model can capture both instantaneous conditions and recent environmental transitions. Using the IoT Agriculture 2024 smart greenhouse dataset, the evaluation compares several Hybrid SVM-LSTM and LSTM-only configurations under proxy labels for actuator-relevant abnormal states. The refined water-related proxy achieved 0.979 accuracy, 0.763 precision, 0.690 recall, and a 0.725 F1-score with 1.142 ms P95 inference time. The broader NPK-gap proxy achieved 0.983 accuracy, 0.886 precision, 0.549 recall, and a 0.678 F1-score with 1.448 ms P95 inference time. These results indicate that the proposed Hybrid SVM-LSTM model is sufficiently compact and accurate for real-time anomaly detection at the edge, especially when proxy-label design and threshold selection are aligned with the target greenhouse condition.
Co-Authors Atam Rifa'i Sujiwanto Seno Adi Putra