DATA GROUPING TECHNIQUES’ PERFORMANCE ANALYSIS IN GM(1,1)’s PREDICTION ACCURACY IMPROVEMENT FOR FORECASTING TRAFFIC PARAMETERS

Vincent B. Getanda, Hidetoshi Oya, Tomohiro Kubo, and Yosuke Sato

Keywords

Performance analysis, number of groups, group data points, dataoverlapping, strong grouping technique, weak grouping technique∗ The Department of Electrical and Electronic Engineering,School of Electrical, Electronic and Information Engineer-ing, Jomo Kenyatta University of Agriculture and Technology(JKUAT), Nairobi, Kenya; e-mail: getandav@kisiipoly.ac.ke∗∗ The Department of Electrical and Electronic Engineering,The Graduate School of Advanced Technology and Science,Tokushima Un

Abstract

This paper is an extended version of our earlier work in which we improved the prediction accuracy of the conventional GM(1,1) by adopting a data grouping technique. We proposed various data grouping techniques and in this paper these data grouping techniques are graphically analysed, and their performance is evaluated in detail. In particular, the strong and weak grouping techniques are discussed, and we show that prediction accuracy continues to improve as the number of data groups increases. In addition, to show that traffic flow characteristics vary as hourly pattern, we forecast vehicle volume and CO2 emissions up to 25 data points unlike in our earlier work in which case we forecasted only 17 data points. Consequently, we unveil more information (e.g. morning peak hour of the day) to discern the happenings of a traffic system. Henceforth we show that 7:20 am is the morning peak time of traffic congestion. Moreover, capturing a longer period of traffic flow scenario validates the concept of data grouping technique.

Important Links:



Go Back