Mean Local Trend Error and Fuzzy-Inference-Based Multicriteria Evaluation for Supply Chain Demand Forecasting

To overcome the inefficiency arising from the separate use of conventional forecast accuracy measures that suffer from the bullwhip effect, especially in uncertain and vague supply chain environments, a forecast accuracy measure, Mean Local Trend Error (MLTE) and a fuzzy-inference-based multicriteria evaluation method are proposed. In contrast to conventional measures, MLTE survives the bullwhip effect by evaluating forecasts based on local trend error. The proposed evaluation method applies fuzzy inference to deal with the uncertainty and vagueness in supply chains and makes a comprehensive evaluation by using an aggregated forecast accuracy index (ACCURACY), which is developed based on fuzzy inference by integrating the proposed MLTE and a conventional measure MAPE, thereby enhancing its efficiency for evaluating supply chain demand forecasts. The proposed MLTE and evaluation method are confirmed by comparative experiments with MAPE based on evaluating four typical forecasting methods – a simple moving average, single exponential smoothing, autoregressive, and autoregressive moving average – on an actual manufacturing-order dataset. The results show that MLTE yields a triple and ACCURACY a quadruple improvement in terms of average distinguishability compared to MAPE. The proposal has potential applications in stock market forecast evaluations.