Weather based paddy yield prediction using machine learning regression algorithms

Abstract

Paddy is a major crop in India which is highly affected by the weather variables resulting in drastic reduction of its yield; adverse all the variables drastically reduce the paddy yield. In this research, machine learning model was developed for prediction of paddy yield production by linear regression (LR), random forest regression (RFR), support vector regression (SVR), cat boost regression (CBR), and hybrid machine learning with variance inflation factor (VIF) LR-VIF, RFR-VIF, SVR-VIF, and CBR-VIF techniques. The dataset consists of variables (weather) for more than 15 years collected for the study area which is Madurai district, Tamil Nadu in India. Analysis was carried out by fixing 70% of data calibration & remaining 30% for validation in Jupyter notebook (Python programming). Results showed that CBR-VIF performed having nRMSE value 1.23 to 1.40% for Madurai South, nRMSE value 0.56 to 1.40% for Melur, nRMSE value 1.10 to 1.25% for Usilampatti, and nRMSE value 0.75 to 1.10% for Thirumangalam. The hybrid model of CBR along with VIF and then CBR model has shown improvement with high influenced weather variables such as maximum temperature, minimum temperature, rainfall normal, and actual rainfall.

Keywords

• Paddy seed
• Hybrid machine learning model
• Linear regression (LR)
• Random Forest regression (RFR)
• Support vector regression (SVR)
• Cat boost regression (CBR).