Stochastic Methodology for Coffee Crop Yield Prediction based on Abiotic Factors

Predicting coffee crop yields improves agricultural decisions, food security, and environmental effect. In agricultural system models, climatic and pedagogical factors affect coffee yield estimates. India produces numerous Arabica and Robusta coffees, two of 103 types. Few countries like Africa, Zimbabwe, Mexico, and Vietnam have studied plantation crop coffee despite its importance. Stochastic machine learning regression models predict coffee output and quantify great abiotic influences using multivariate feature selection technique and feature importance methods which selects the best features based on multivariate statistical tests. Our research used Karnataka's Central Coffee Research Institute (CCRI) coffee research station's 2004–2022 abiotic factors: year, rainfall, temperature, sunlight, RH, vapour, dew points, and yield. Abiotic predictor variables were used in the present two-model research on elastic net and extra tree models. The features have been grouped into group-1, 2, 3, 4, and 5 based on the multivariate feature selection technique and feature importance methods. Group-1 features had the highest constant of determination for the 70:30 splitting proportion, using Year, Rainfall, Temperature, Sunshine, Relative Humidity, Vapour, and Dew point as predictor variables (Rsquare = 0.72 and Root Mean Square Error = 76.54 kg per ha) for Elastic net model and coffee yield R-square = 0.53 and Root Mean Square Error = 96.75 kg per ha was the best coefficient of determination for the extra tree regression model with group-3 parameters (Year, Year + Relative Humidity Minimum and Maximum + Rainfall + Minimum and Maximum Temperature). This research found the best coffee production weather parameter, defeating extra tree model.