Integrating Fuzzy Rule Based Reasoning with Machine Learning for Intelligent and Scalable Irrigation Control

Bharati Wukkadada; Pradeep Kachakayala; padma mishra; Aditi Srivastava; T. Prem Chander; Saiprasad Potharaju; Ali Khwaldeh

doi:10.37965/jait.2026.1853

Integrating Fuzzy Rule Based Reasoning with Machine Learning for Intelligent and Scalable Irrigation Control

Authors

Bharati Wukkadada Department of Data Science and Technology, K J Somaiya Institute of Management, Somaiya Vidyavihar University, Mumbai, India
Pradeep Kachakayala Cloud Engagement Manager, Insight Enterprises, Arizona, USA
padma mishra Master of Computer Applications, Thakur Institute of Management Studies, Career Development and Research, Maharashtra, India
Aditi Srivastava Master of Business Administration, ABES Business School, Ghaziabad, India https://orcid.org/0000-0003-0035-3108
T. Prem Chander Department of CSE, Matrusri Engineering College, Saidabad, Hyderabad, Telangana, India
Saiprasad Potharaju Computer Science and Engineering, Symbiosis Institute of Technology, Symbiosis International (Deemed University), Pune, India https://orcid.org/0000-0002-7511-6855
Ali Khwaldeh Communications and Computer Engineering Department, Hourani Center for Applied Scientific Research, Al-Ahliyya Amman University, Amman, Jordan

DOI:

https://doi.org/10.37965/jait.2026.1853

Keywords:

decision support, environmental conditions, fuzzy logic, irrigation, machine learning

Abstract

This study presents a hybrid intelligent irrigation decision support system that integrates fuzzy logic and Random Forest (RF) regression to support irrigation decision-making in agriculture. A fuzzy inference engine was designed using expert-defined rules based on soil moisture, air temperature, and rainfall. This system generated irrigation recommendations on a scale of 0 to 10 using Mamdani-type fuzzy logic. An RF model was trained on the same inputs with the fuzzy output as the target to evaluate how accurately a data-driven model can reproduce the fuzzy decision mapping. Using five-fold cross-validation, the RF achieved a root mean squared error (RMSE) of 0.2459 , mean absolute error (MAE) of 0.0801 , and coefficient of determination (R2) of 0.9682, indicating strong alignment with fuzzy decision logic. This hybrid approach offers a transparent, data-driven, and scalable solution, effectively bridging the gap between interpretability and prediction accuracy. The system has the potential to support precision irrigation and informed water management decisions and was evaluated using one year of daily aggregated Internet of Things (IoT) sensor observations from Narrabri, Australia.