Predictive modelling of apple area, production, and productivity in Srinagar

This study evaluates and compares the performance of linear and non-linear regression models to analyse trends and predict key metrics-namely area, production, and productivity-of apple cultivation in Srinagar over a 25-year period (1995 to 2019). Four models were examined: Linear, Logistic, Monomolecular, and Exponential. Their goodness of fit was assessed using multiple statistical indicators, including R², Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Akaike Information Criterion (AIC), the Run Test, and the Shapiro–Wilk Test. The findings reveal that non-linear models, especially Logistic and Monomolecular, outperform the linear model with higher R² values and lower error metrics across all parameters. The Logistic model proves most effective for estimating area and productivity, whereas the Monomolecular model excels in forecasting production. Furthermore, the comparison between actual and estimated values demonstrates the robustness of these models in capturing historical patterns and predicting future changes. Specifically, the Logistic model provides reliable estimations for maximum area and productivity with minimal deviations from observed data, while the Monomolecular model captures production trends with high precision. These results offer valuable insights for policy planning and for promoting sustainable apple cultivation practices in the region. Future research should explore integrating environmental and economic variables to further refine these models.