DATA-DRIVEN FORECASTING MODEL FOR SMALL DATA SETS

The effective use of information is an important foundation for a company's sustainable development and effective management of its operations. However, when making important management decisions, managers often face issues of insufficient information or limited data due to time or cost constraints. The use of a grey model is a common solution for solving this small-data-set issue; this model has been successfully applied to various fields with reliable outcomes. Nevertheless, it may not always achieve sufficient accuracy, especially in predicting non-equigap data. This study introduces the concept of fuzzy membership functions to reform the formula of the background values of the data that is input, and then proposes a data-driven grey model for a small amount of non-equigap data. Two real case studies involving material-fatigue-limit testing data and the monthly demand for a specific uninterruptible power supply product are taken as examples to demonstrate the proposed method. The experimental results show the proposed method is able to obtain a solid outcome, yielding accurate forecasts using a small amount of non-equigap data.