Date of Award


Publication Type

Master Thesis

Degree Name



Computer Science

First Advisor

Yuan, Xiaobu




The rapid development of urban areas and the increasing size of vehicle fleets are causing severe traffic congestions. According to traffic index data (Tom Tom Traffic Index 2016), most of the larger cities in Canada placed between 30th and 100th most traffic congested cities in the world. A recent research study by CAA (Canadian Automotive Association) concludes traffic congestions cost drivers 11.5 million hours and 22 million liters of fuel each year that causes billions of dollars in lost revenues. Although for four decades’ active research has been going on to improve transportation management, statistical data shows the demand for new methods to predict traffic flow with improved accuracy. This research presents a hybrid approach that applies a wavelet transform on a time-frequency (traffic count/hour) signal to determine sharp variation points of traffic flow. Datasets in between sharp variation points reveal segments of data with similar trends. These sets of data, construct fuzzy membership sets by categorizing the processed data together with other recorded information such as time, season, and weather. When real-time data is compared with the historical data using fuzzy IF-THEN rules, a matched dataset represents a reliable source of information for traffic prediction. In addition to the proposed new method, this research work also includes experiment results to demonstrate the improvement of accuracy for long-term traffic flow prediction.