Machine learning and Sensor-Cloud Based Precision Agriculture for Intelligent Water Management for Enhanced Crop Productivity

Abstract

The combination of Machine Learning algorithms with Internet of Things devices is emerging as an effective solution to redefining precision agriculture for better water management and crop cultivation. The purpose of this study is to use different learning models, such as Artificial Neural Networks , Support Vector Machines , Decision Trees , and Random Forest , to predict the irrigation need based on real-world sensor data. To retrieve the output variable, which is the irrigation requirement, data from temperature, soil moisture, water level, and humidity sensors that are available in a tomato cultivation facility are used. The dataset consists of 3422 readings, which are split into training and testing sets. A designated percentage of 70% is used for training the models, while the remaining 30% are used to test the outcomes. As the results of the study show, the ANN model is the most accurate predictor of irrigation need with the classification rate of 97.6%, followed by SVM , DT , and RF with 95.4%, 91.3%, and 88.9% correspondingly. The differences in the outcomes are demonstrated in confusion matrices, which identify the classification of the cases and indicate the percentage of correct predictions. Evidence of the predictive power of ML models implies that farmers can independently determine when to activate the pump when real-world data serve as the input. Additionally, the ability to collect real-world data using IoT sensors is beneficial, as it empowers farmers with up-to-date information to make a timely decision about pump activation. The limitations are associated with the type of crop and agricultural facility and, for this reason, future studies may investigate the generality of the conclusion with regard to other crop types and facilities.