The main objective of this study is to evaluate BFTSC (break for time series components) and GFTSC (group for time series components) identification of time series components. The weaknesses of BFAST (Break for Additive Seasonal and Trend) were corrected by the extension of BFAST to BFTSC which resulted into creation of two new technique named Break for Time Series Components (BFTSC) and Group for Time Series Components (GFTSC).BFTSC is created to capture the trend, seasonal, cyclical and irregular components as a combined image and to present them in a single plot. Group for Time Series Components (GFTSC) is designed to capture all the time series components on a different individual time plot. BFAST only identifies trend and seasonal components while considering all others as random. BFTSC and GFTSC is created to include cyclical and irregular components and this was included in the methodology. Evaluation using simulation data was conducted to verify the accuracy of BFTSC and GFTSC. For yearly sample size of 8 years small sample, 16 years medium sample, 24 years large sample. For the monthly data, 48 months small monthly sample size, 96 monthly medium sample size, 144 months large sample size. Each of the sample size was replicated 100times each to form a total of 163,200 data unit. GFTSC outperform BFTSC With0.2% while both GFTSC and BFTSC are effective and better than BFAST because it was able to identify 100% of the data with the basic four time series components monthly. BFTSC detects 99% of the entire components in the time series monthly data that was tested while GFTSC identify 99.2 of the entire components in the time series monthly data that was tested. Subsequently the two methods can be used to determine the next forecasting technique and can provide a better alternative to BFAST technique.
Volume 11 | 05-Special Issue
Pages: 995-1004