Optimizing Production Planning in Cement Manufacturing Using SVR

The devastating earthquake that caused significant destruction in 11 provinces on February 6, 2023, has accelerated the growth rate of the cement sector. This rapid growth, coupled with increasing stock market activity, marks a golden age for the sector while emphasizing the critical need for accurate future production forecasting. Leveraging the predictive capabilities of machine learning algorithms, experiments were conducted using five years of production data from a cement factory in the Southeastern Anatolia region. The Support Vector Regression (SVR) model, an application of the Support Vector Machine (SVM) algorithm, was tested with RBF, linear, sigmoid, and polynomial kernels. Among these, the SVR model with the RBF kernel yielded the best performance across four evaluation metrics: Mean Squared Error (MSE): 0.002926, Root Mean Squared Error (RMSE): 0.054094, Mean Absolute Error (MAE): 0.048611, and Mean Absolute Percentage Error (MAPE): 0.052697. This paper highlights the effectiveness of SVR-RBF in providing reliable production forecasts for the cement industry and supporting strategic planning to address dynamic market demands.