Modeling and forecasting of milk production in different breeds in Turkey

Abstract

Identification of milk production is one of the key activities for the Turkish economy in terms of providing economic income for dairy smallholders and meeting consumer demands. This study aims to predict milk production in Turkey using various time series models which are BATS, TBATS, Holt's Linear Trend, ARIMA models, and NNAR. Yearly data from 1991 to 2019 on Milk Production is used in this study. The best time series model is selected from the testing data set (2015 to 2019) based on the lowest Mean Absolute Percentage Error (MAPE). The results of this study showed that the best predicts are obtained for Culture purebred milk production by an ARIMA (1,2,1) model, for Crossbreed milk production by a TBATS Model, for Indigenous milk production by ARIMA (0,2,0) model for total milking cows' production by Holt’s Linear Trend model. Furthermore, these models forecasted an increasing trend in the production of milk from 2020 to 2025. The percentage increases for culture purebred milk, Crossbreed milk production, and Indigenous milk production from 2020 to 2025 are projected to be 40, 20.9 and 10.9%, respectively. Overall, the total milking cows' production is projected to increase by 25% in 2025.

Keywords

• ARIMA
• Forecasting
• Holt’s Linear Trend
• Milk production
• TBATS