Comparison of estimation methods in univariate estimation: The Turkish cement industry case

This study aims to develop a method for forecasting cement production in the Turkish cement manufacturing industry and to draw conclusions about future production development by applying it. The data used in this study is the total monthly cement production of the Turkish cement industry. The data consists of 93 months, from January 2017 to September 2024. Artificial neural network modelling procedures were applied to the model, and forecasting was performed using deep learning tools. The estimation's performance was quite good, and it was concluded that the network could be generalised. The results showed that the model developed can forecast monthly production, with threshold values and weights obtained from the trained network. Monthly cement forecasts were issued through December 2026. The success of modelling univariate time series data with ANNs was validated by comparing it with Winters' and SARIMA methods using MAPE statistics. The most suitable forecasting method was found to be ANNs, and the findings were discussed.