Optimizing Irrigation Efficiency with IoT and Machine Learning: A Transfer Learning Approach for Accurate Soil Moisture Prediction
Srinivasa Rao Burri, Deepak Agarwal, Narayan Vyas, Ronak Duggar
Abstract
This research aims to develop a Machine Learning model for predicting soil moisture levels, which may be used to construct smart irrigation systems. The model was evaluated and trained using data from the “Smart Irrigation System Dataset” made publicly available by the University of California, Irvine. A transfer-learned ResNet50 model is evaluated using various classification measures like accuracy, recall, precision, and area under the ROC curve (AUC). The proposed model has an AUC of 0.95, meaning it correctly identifies positive and negative samples 95% of the time. Moreover, the model’s performance is measured against that of other famous machine learning models like logistic regression, Support Vector Machines (SVM), K-Nearest Neighbours (KNN), random forests, decision trees, and naive Bayes, with the majority of these conventional models being outperformed. These findings have ramifications for researchers and engineers creating intelligent irrigation systems for precision agriculture.