Gourari et al. (2026) An artificial intelligence-based stacking ensemble framework for smart irrigation pump control using IoT sensor data

Identification

• Journal: International Journal of Power Electronics and Drive Systems/International Journal of Electrical and Computer Engineering
• Year: 2026
• Date: 2026-09-18
• Authors: Sarra Gourari, Wafa Difallah, Belkacem Draoui
• DOI: 10.11591/ijece.v16i5.pp2652-2663

Research Groups

• Laboratory of Information Processing and Telecommunication (LTIT), Department of Mathematics and Computer Science, Faculty of Exact Sciences, Tahri Mohammed University of Bechar, Bechar, Algeria
• Innovation in Informatics and Engineering Laboratory (INIE LAB), Department of Mathematics and Computer Science, Faculty of Exact Sciences, Tahri Mohamed University, Bechar, Algeria
• Laboratory of Energetic in Arid Zones (ENERGARID), Faculty of Technology, Tahri Mohammed University of Bechar, Bechar, Algeria

Short Summary

This study proposes a stacking-based framework for predicting irrigation pump operation in an ON/OFF classification setting. The framework uses environmental and soil-related variables to combine the outputs of three base learners: Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Multilayer Perceptron (MLP). The proposed ensemble achieved 99.97% accuracy, 99.94% precision, 100.00% recall, 99.94% specificity, and a 99.97% F1-score.

Objective

• Develop an intelligent irrigation decision-support system that can predict pump operation (ON/OFF) based on environmental and soil sensor data using a stacking ensemble framework.

Study Configuration

• Spatial Scale: Local scale, focusing on individual irrigation pumps.
• Temporal Scale: Real-time prediction of pump operation status.

Methodology and Data

• Models used:
  • Random Forest (RF)
  • Extreme Gradient Boosting (XGBoost)
  • Multilayer Perceptron (MLP)
  • Logistic Regression (meta-learner)
• Data sources:
  • IoT sensor data from smart irrigation systems
  • Publicly available datasets from Kaggle

Main Results

• The proposed stacking ensemble achieved 99.97% accuracy, 99.94% precision, 100.00% recall, 99.94% specificity, and a 99.97% F1-score.
• The individual models (RF, XGBoost, MLP) achieved accuracies ranging from 97.34% to 99.95%.
• The stacking ensemble outperformed the strongest individual base learner (XGBoost) by improving accuracy by 0.02 percentage points and F1-score by 0.01 percentage points.

Contributions

• This study addresses the limitations of previous studies on stacked ensembles for binary irrigation pump-status prediction.
• The proposed framework combines the strengths of multiple machine learning models to improve predictive performance.
• The study provides a deployment-oriented evaluation, reporting per-model inference latency, throughput, and serialized memory footprint.

Funding

• This research was funded by [insert funding agency or project name].

Citation

@article{Gourari2026artificial,
  author = {Gourari, Sarra and Difallah, Wafa and Draoui, Belkacem},
  title = {An artificial intelligence-based stacking ensemble framework for smart irrigation pump control using IoT sensor data},
  journal = {International Journal of Power Electronics and Drive Systems/International Journal of Electrical and Computer Engineering},
  year = {2026},
  doi = {10.11591/ijece.v16i5.pp2652-2663},
  url = {https://doi.org/10.11591/ijece.v16i5.pp2652-2663}
}