A grey silhouette coefficient for the small sample forecasting

Small sample forecasting problem is an important issue in various fields. The early stage of manufacturing system is a positive example about this issue. Manufacturers need sufficient management knowledge to lower overall production cost, but it is a hard task due to the obtained samples is limited. This study is thus to develop a modeling procedure to acquire stable prediction results under small data sets. Briefly, we first judge some single models to determine whether the real sequence tendency can be reflected with the grey incidence analysis and then evaluate their forecasting stability by the relative ratio of error range; finally, the grey silhouette coefficient is developed to build an applicable hybrid forecasting model for small samples. The material fatigue limit data set is used here to confirm the effectiveness and practical application value of the proposed method. The empirical results show that the hybrid model indeed can lower forecasting errors and come up better results with the limited data. Consequently, the proposed procedure is considered a feasible tool for the small sample forecasting problem.