02. Logistics and Supply Chain Management
As the information that people could obtain is most of the times limited and fell under the category of uncertainty circumstances, a variety of uncertainty system theories, such as fuzzy sets (FS), rough sets (RS) and grey system theory (GST), emerged conforming to the requirements of the new era. The GST is considered as a multi-disciplinary theory that deals with systems that lack information. GST is a novel system theory which mainly focuses on analysing and modelling problems with limited data or poor information (Xie, 2017). This is one of the characteristics that prevail in the supply chain. Not all the information nor its reliability is always possible. All these models have solved plenty of practical application problems, such as economic forecasting, management decision making, scientific evaluation, energy management, industrial process control, water resource analysis, risk management and agricultural breeding research (Lin, et al., 2004). Surprisingly, none of these models have been applied in the supply chain area. A Grey system is a system containing grey inputs and grey or regular outputs. Grey information could be further explained as limited valuable sequence data, grey number information and grey system structures. A grey model based on limited data is a new type of forecasting model for solving prediction problems based on limited valuable sequence data. In traditional statistical forecasting models, data must meet the requirement of scale, i.e. large scale of data should be collected. This condition could not always be satisfied. This is not always possible when you plan to forecast the demand in supply chain. As a result of this investigation, I have found that this model can obtain better prediction results, especially for data of small sample sizes (<20 records). The paper presents a methodology for incorporating limited or incomplete data into a modified GM (1,1) model applied in the supply chain area. The results obtained exceed other models and methodologies such as: linear regression, moving averages, exponential smoothing and are applied where ARIMA is not possible, since the stationary condition in GST is not required for data samples below 20 records. The model also identifies the algorithm to identify when the GM (1,1) is applicable: (i) limited amount of data (<20 records), (ii) stationary is not present, (iii), no seasonal or cyclical demand is found, and (iv) forecast projections are required for periods >n+1. The evaluation methodology was implemented by using Minitab and R. Metrics for calculating the forecasting error and evaluating its performance are: (i) forecast error, (ii) mean absolute error (MAE), (iii) mean square of error (MSE) and (iv) mean absolute percentage error (MAPE) (Vagale, et al., 2021). The results by using the GM (1,1) are better than the results of the linear regression and moving averages, 1.6% vs 5.7% respectively. The results show that the prediction ability of the grey prediction with GM (1,1) model is better than traditional approach.
PALAVRAS-CHAVE: forecasting, grey systems, grey markov, supply chain.