Event Index Computation for Forecasting Case Study: Car Sales in Thailand

Witchaya Rattanametawee, Chartchai Leenawong

Abstract

Due to the impact of special events, both positive and negative, on the sales data, the ordinary Time-series Decomposition (TSD) forecasting model cannot merely capture these effects, even with the added seasonality and trends. Therefore, in this research, a new method for computing the event indices, representing the unusual fluctuations for a certain period in the time series, is proposed in order for it to be incorporated into TSD, alongside the conventional trend, seasonal, and cyclical components. A case study of subcompact car sales monthly data in Thailand during the years 2011-2018 is examined as for that time period contains the 2011 nationwide big flood reflecting the negative impact, as well as the nation's tax-incentive first-car buyer scheme reflecting the positive impact on the dataset. The mean absolute percentage error (MAPE) is used as an accuracy measure of the proposed forecasting model and it illustrates the promising results in the end.