A comparative study of time series and wavelet-based models for accurate Mesta price forecasting in Andhra Pradesh

This study focuses on comparing traditional time series models and wavelet decomposition-based models to identify the most effective method for forecasting Mesta wholesale prices in the Amadalavalasa market of Andhra Pradesh. Utilizing monthly price data from January 2010 to December 2024, various models including ARIMA, GARCH, Artificial Neural Networks (ANN), Wavelet-ARIMA (WARIMA), Wavelet-ANN (WANN), and Wavelet-GARCH (WGARCH) were evaluated. Model performance was assessed using accuracy metrics such as Root Mean Squared Error (RMSE), Mean Squared Error (MSE), and Mean Absolute Percentage Error (MAPE) on both training and testing datasets. The results demonstrated that wavelet decomposition-based models, particularly WARIMA, outperformed traditional models by providing more balanced and consistent forecasts with strong generalization capabilities. While ARIMA exhibited stable and reliable predictive accuracy, other models like GARCH and wavelet-based nonlinear approaches showed less consistent performance. The findings underscore the advantage of incorporating wavelet decomposition in time series forecasting to capture complex price dynamics, thereby improving the accuracy and reliability of Mesta price predictions for market stakeholders.