A Comparative Performance Evaluation of Standalone and Hybrid Forecasting Models for Cashew Nut Price Prediction in Andhra Pradesh, India

Abstract

Cashew nut (Anacardium occidentale L.) is an important plantation and export crop that plays a significant role in the livelihoods of farmers and the agricultural economy of Andhra Pradesh. Accurate price forecasting is essential for better market planning and informed decision-making in the cashew sector. The present study compared standalone and hybrid forecasting approaches for predicting cashew nut prices. Autoregressive Integrated Moving Average (ARIMA) and Support Vector Regression (SVR) models were developed individually, while a hybrid ARIMA–SVR model was used to capture both linear and nonlinear movements in the price series. The performance of the models was assessed using Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE). Although the SVR model recorded lower error values, its residuals showed significant autocorrelation, suggesting that the model could not fully capture the underlying pattern of the series. Hence, the hybrid ARIMA–SVR model, which was the next best-performing model, was considered more suitable for forecasting cashew nut prices. Further diagnostic tests such as the Box–Pierce, BDS, and Diebold–Mariano tests also supported the adequacy and predictive superiority of the hybrid model. Forecasts for the period from January to June 2026 indicated a slight declining trend in cashew nut prices. Overall, the hybrid ARIMA–SVR model provided more reliable forecasts and can support better market-related decisions in Andhra Pradesh.