Forecasting accuracy analysis based on two new heuristic methods and Holt-Winters-Method

摘要

Since 1970s, many academic researchers and business practitioners have started to develop different forecasting methods and models. Most of them are still used in the IT-Systems nowadays. However, they don't perform well enough in practice. People pay much attention to data collection but ignore the data quality, which could lead to low forecasting accuracy. In this paper, we will introduce two new heuristic business forecasting techniques (Revinda and Metrix). Both methods utilize inherent structures of time series. The error analysis is based on B2C and B2B aggregated commercial data. In addition, these two methods will be compared with HoLT-WiNTERS-Methods (HWM) by using error measures MAPE, percentage better and Theil's U2.