Evaluating the machine learning based Efficacy of Decision Tree and Support Vector Machines in Smart Irrigation Systems for Precise Irrigation Status Classification for Optimizing Water Management in Agriculture

Abstract

The research paper deals with the improvement of water management practices in agriculture through the use of machine learning in smart irrigation systems. The efficiency of the Decision Tree and the Support Vector Machine models with real-time sensor data was analyzed. The Internet of Things system was developed to analyze the data collected at the field with the help of various sensors, including temperature, humidity, and soil moisture sensors. The layout and choice of sensors were based on the official parameters required for most average plants and corps. After data collection and development of classifiers, the Decision Tree and Support Vector Machine models were used and analyzed regarding their efficiency for making irrigation-related decisions. As the result, it was found that both models showed a good performance, but the SVM model was slightly better due to a smaller number of false positives and false negatives, which allow it to divide the data into the corresponding categories more accurately. At the same time, both models use history data, and the more recent data are better for decision as it was also noted during the development when the use of models based on history data for irrigation decision result in regular irrigation, which in long term would cause the loss of productivity of plans, due to the issue that ancient and no crops are watered too. As the final results, it can be stated that the models show good performance, and the algorithms that were tested can be stated viable for the use at practice. At the same time, the SVM model shows slightly better performance than DT.