Date of Award


Publication Type

Master Thesis

Degree Name



Computer Science

First Advisor

Yuan, Xiaobu




It is essential for intelligent transportation systems to be capable of producing an accurate forecast of traffic flow in both the short and long terms. However, the counting datasets of traffic volume are non-stationary time series, which are integrally noisy. As a result, the accuracy of traffic prediction carried out on such unrefined data is reduced by the arbitrary components. A prior study shows that Box-Jenkins’ Autoregressive Integrated Moving Average (ARIMA) models convey demand of noise-free dataset for model construction. Therefore, this study proposes to overcome the noise issue by using a hybrid approach that combines the ARIMA model with fuzzy wavelet transform. In this approach, fuzzy rules are developed to categorize traffic datasets according to influencing factors such as the time of a day, the season of a year, and weather conditions. As the input of linear data series for ARIMA model needs to be converted into linear time series for traffic flow prediction, the discrete wavelet transform is applied to help separating the nonlinear and linear part of the time series along with denoised time series traffic data.