A Light gradient boosting machine regression model for prediction of agriculture insurance cost over linear regression

Abstract

To increase accuracy for the prediction of agriculture insurance claim cost based on crop insurance data.Gradient Boosting Machine (LGBM) and linear regression models are tested with total Samples 6022 for n=7 iterations to predict accuracy. LGBM works based on decision tree algorithm and linear based on fitted regression equation. The coefficient of determination values of proposed LGBM regression (92.52%) and linear regression (72.47%) are obtained. There was a statistical significance between LGBM regression and linear regression (p=0.001).Prediction of agriculture insurance claim cost LGBM regression technique produces significantly better performance than the linear regression technique.