الفهرس 
AbstractOnly 14 pages are availabe for public view
 |  | from | 238 |  |  |
| | | from | 238 | | |
Abstract
Traffic congestion is an important socio-economic problem that swelled in the last few decades. It affects the social mobility of people, length of trips, quality of life, and the economy of countries. As a major problem in most countries, it has been tackled by governments, universities, and advanced research using intelligent transportation systems (ITS) to solve the problem or at least ease its adverse effects. Intelligent Transportation Systems (ITS) has become the cutting edge solution to most traffic problems. One of the important problems is the prediction of the incoming traffic pattern on the highway. There is a number of available ITS approaches for traffic congestion prediction. In this work, the non-deterministic Adaptive NeuroFuzzy Inference System (ANFIS), the statistical Hidden Markov Models (HMM), and a hybrid HMM-ANFIS method are used for short term traffic prediction. In the ANFIS method, two variables are used as inputs. Those are speed and density, each with three input levels. The speed levels are slow, medium and fast while the density levels are low, medium and high. The output of the ANFIS model is the Level of Congestion (LOC). It has nine different congestion levels ranging from free-flow traffic to serious congestion. The results of this approach showed 15% MAE for similar approach to previous research and showed 11% MAE for the modified approach in this work. Another approach using HMM to define the traffic states during peak hours in two dimensional space using mean speed and contrast. The novelty in this research include the use of contrast with HMM is proved to be a good measure for traffic prediction, contrast suitability for short-term prediction is extended to 90 minutes, and the contrast range of values is proved to be different for different traffic datasets. The method showed a classification error of 13% and a prediction error of 8.8%. Lastly, a novel hybrid approach using Hidden Markov Models (HMM) and Adaptive NeuroFuzzy Inference System (ANFIS) is introduced. HMM is implemented to take into consideration time factor. It is used to select the right NeuroFuzzy network suitable for this particular time period for efficient congestion prediction. The novelty in this research include the right choice of traffic pattern for training affects the quality of the prediction dramatically. The results from the hybrid model showing 6% MAE rate which outperforms the standard stand alone neurofuzzy approach of 15% error. The empirical evaluation of the research is based on a UK dataset which is provided by the UK Department of Transport. The data is a yearly statistics from 2009 to date and is available in a monthly ’’.CSV’’ files. It was collected using loop detectors and aggregated every 15 minutes for various parts of the UK motorways. The link under consideration is A453 between A50 and A42 (AL1260). Empirical results show that the proposed approach is effective in traffic prediction compared to related work. Moreover, the research methods are verified using a case study from Egypt to ensure the validity of the algorithms for Egyptian traffic.