<![CDATA[Accurate cement stock prediction is crucial for optimizing supply chain management, ensuring operational efficiency, and achieving long-term sustainability and profitability within the global cement industry. Inaccurate predictions can lead to significant costs due to overstocking or stockouts, impacting customer satisfaction and overall economic development. The complex market environment and dynamic demand fluctuations within the construction sector further exacerbate these challenges. This study compares time series forecasting algorithms to predict cement stock levels. The methodologies investigated include traditional statistical models: Simple Moving Average (SMA), Double Moving Average (DMA), Simple Exponential Smoothing (SES), and Double Exponential Smoothing (Holt's Method). Additionally, advanced machine learning and deep learning models, namely ARIMA (Autoregressive Integrated Moving Average), LSTM (Long Short-Term Memory), and Prophet, are also evaluated. This research aims to identify the most suitable algorithm for cement stock forecasting by assessing their performance using standard metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE). Initial findings from existing literature suggest that while traditional methods offer simplicity, modern models like LSTM often achieve superior accuracy for complex and non-linear patterns, whereas Prophet excels at automatically handling seasonality and missing data.ARIMA provides computational efficiency for simpler, stationary patterns but may struggle with non-linearity. This study contributes to the field by providing a structured comparison of diverse forecasting techniques specifically tailored for cement inventory, offering practical guidance for industry practitioners and informing strategic decision-making in supply chain optimization. ]]>
Copyrights © 2025
