Kumar et al. (2026) A Decomposition-Driven Deep Learning Model for Intelligent Irrigation Decision-Making

Identification

• Journal: International Research Journal of Multidisciplinary Technovation
• Year: 2026
• Date: 2026-09-22
• Authors: Rajesh Kumar, Anil Garg
• DOI: 10.54392/irjmt26512

Research Groups

• Department of Electronics & Communication Engineering, Maharishi Markandeshwar Engineering College, Maharishi Markandeshwar (Deemed to be University), Mullana-Ambala, Haryana, 133207, India

Short Summary

This study proposes a hybrid model that combines Convolutional Neural Network (CNN) with Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) for irrigation classification and water conservation in agricultural fields. The proposed model achieved the highest accuracy (98%), precision (0.95), recall (1), and F1-score (0.97) compared with benchmark models.

Objective

• Investigate the effectiveness of a hybrid CEEMDAN-CNN model for intelligent irrigation decision-making.
• Evaluate the performance of the proposed model using various metrics, including accuracy, precision, recall, and F1-score.

Study Configuration

• Spatial Scale: Local (farm-level)
• Temporal Scale: Real-time

Methodology and Data

• Models used: Convolutional Neural Network (CNN), Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN)
• Data sources: Soil moisture, temperature, humidity data from a YL-69 soil-moisture sensor and DHT-11 temperature and humidity sensor

Main Results

• The proposed CEEMDAN-CNN model achieved an accuracy of 97.83% with a standard deviation of 1.12% and a 95% confidence interval of 97.51–98.15%.
• The mean precision, recall, and F1-score were 96.90%, 97.42%, and 97.16%, respectively.
• The model also achieved ROC-AUC, PR-AUC, balanced accuracy, and MCC values of 99.12%, 98.36%, 97.64%, and 95.27%, respectively.

Contributions

• This study introduces an integrated CEEMDAN-CNN framework for irrigation classification, which leverages CEEMDAN-derived multiscale feature representations to capture hidden temporal and frequency-dependent characteristics of soil moisture, temperature, and humidity signals.
• The proposed approach is systematically evaluated against conventional ML, DL, and other decomposition-based frameworks.

Funding

• This research was funded by the Department of Electronics & Communication Engineering, Maharishi Markandeshwar Engineering College.

Citation

@article{Kumar2026DecompositionDriven,
  author = {Kumar, Rajesh and Garg, Anil},
  title = {A Decomposition-Driven Deep Learning Model for Intelligent Irrigation Decision-Making},
  journal = {International Research Journal of Multidisciplinary Technovation},
  year = {2026},
  doi = {10.54392/irjmt26512},
  url = {https://doi.org/10.54392/irjmt26512}
}