Özet:

Abstract Timely and accurate crop yield prediction serves as a pillar for the country’s food security and frames the strategic policies for the government. In this study, we endeavoured to assess the effectiveness of three various machine learning-based methods to predict paddy yield for the Indian state of Andhra Pradesh. The models were developed using historical yield data for the years 2001 to 2020 along with the long-term derived satellite variables evapotranspiration (ET), leaf area index (LAI), land surface temperature (LST), normalised difference vegetation index (NDVI), and rainfall (RF). Multiple linear regression (MLR), support vector regression (SVR), and random forest regression (RFR) models were three different machine learning models that were assessed for performance. A correlation was established between these variables and crop yield. The highly correlated features model was built and the features with the least correlation were discarded. The performance of all three models was found to be satisfactory. The RFR model was found to have higher accuracy with an R2 value of 0.61 and an RMSE of 0.55 t ha-1. Whereas MLR and SVR were found to have R2 0.51 and 0.59, RMSE 0.59 t ha-1, and 0.54 t ha-1. The results from the current study have shown the capability of machine learning algorithms with limited datasets.

Anahtar Kelimeler: