Comparative analysis of time series models for forecasting raspberry production in Serbia

The aim of this paper is to model and forecast raspberry production in Serbia based on annual time series data for the period from 1970 to 2024. In the analysis, Exponential Smoothing models (Brown and Holt-Winters) and Autoregressive Integrated Moving Average models are applied, and compared to determine the most appropriate econometric model to describe and forecast the development of raspberry production in Serbia for the period from 2025 to 2030. The performance of the model was evaluated based on several criteria: Root Mean Square Error, Mean Absolute Error and Mean Absolute Percentage Error, both in-sample (2016-2020) and out-of-sample (2021-2024). Based on the observed prediction accuracy of the selected models, it can be concluded that the ARIMA model provides the best results for the prediction of raspberry production. According to the ARIMA model, a slight downward trend in raspberry production is expected from 2025 to 2030.