A Demand Forecasting Result Comparison Between Make-to-Stock and Make-to-Order Manufacturing: Using a Hybrid ARIMA-LSTM Model

Demand forecasting plays a crucial role for manufacturing companies in optimizing resource including materials, machinery, manpower, production, and logistics planning. In recent years, within the VUCA (volatility, uncertainty, complexity, ambiguity) context, manufacturers have had to contend with the fluctuating customer demand. Capturing consumer demand for products and services before starting production activities is essential in every business. Both make-to-order (MTO) and make-to-stock (MTS) manufacturing face significant challenges in forecasting due to unpredictable demand and the rapid advancement of cutting-edge technologies. Achieving faster, more accurate, and real-time demand forecasting is a significant goal for businesses. Therefore, this paper proposes a hybrid ARIMA-LSTM model. While the ARIMA model can only process and forecast univariate and linear time series data, the LSTM model is capable of handling nonlinear patterns. Our proposed method combines the advantages of both ARIMA and LSTM models, facilitating the processing of high volumes and complex time series data. Experimental results demonstrate that our proposed method outperforms the individual ARIMA and LSTM models in terms of performance measures.