Comparative Analysis of Machine Learning Algorithms for Irrigation Status Prediction

Abstract

Compared to traditional methods, systems supported by machine learning have been found to make more accurate irrigation decisions. In this study, Kaggle, a data sharing platform frequently used in data science, was utilized. The “Weather Data” dataset was used for irrigation status prediction. Missing and outlier data were corrected and dependent (irrigation status) and independent (air temperature, humidity, soil moisture value, precipitation) parameters were obtained. Focusing on the provinces in the Southeastern Anatolia Region (Adıyaman, Batman, Diyarbakır, Gaziantep, Kilis, Mardin, Siirt, Şanlıurfa and Şırnak), the accuracy of the algorithms was tested under different conditions. In the analyses conducted separately for each province, irrigation status estimation was performed using machine learning algorithms such as Decision Trees, Support Vector Machines, Random Forest, Naive Bayes, Gradient Boosting, Logistic Regression, K-Nearest Neighbor and Artificial Neural Network models. As a result of the predictions made on the dataset, the algorithms were compared using the accuracy metric and the most effective algorithms were Random Forest (95%), Decision Tree (97%), Gradient Boosting (93%) and Artificial Neural Network (98%) models with over 90% accuracy in all cities. Other algorithms also showed remarkable performance with high accuracy rates (above 75%). In the analysis for each city, the algorithm performance ranking was similar. In conclusion, this study shows that machine learning algorithms can be used in agricultural irrigation with very high performance.