DSpace Repository

Application of neural network and fuzzy inference system for bus service quality prediction and attribute ranking in Dhaka city

Show simple item record

dc.contributor.advisor Hadiuzzaman, Dr. Md.
dc.contributor.author Rokibul Islam, MD.
dc.date.accessioned 2017-07-25T07:15:35Z
dc.date.available 2017-07-25T07:15:35Z
dc.date.issued 2016-11
dc.identifier.uri http://lib.buet.ac.bd:8080/xmlui/handle/123456789/4559
dc.description.abstract With a view to assessing the service quality (SQ) provided by the bus transit system, this study has conducted field survey through face-to-face interviews at main bus stops around Dhaka city throughout the month of December, 2014. The questionnaire was based on 22 attributes and some demographic variables. These attributes have been selected by analyzing users’ demand and the transit experts’ view towards service quality indicators. After conducting the on-board and offboard survey, a set of 655 samples has been selected for analyzing with Artificial Intelligence (AI) models. AI is a strong method to simulate the human decisions to assess the quality of services depending on some attributes. However, to get the best tool for bus SQ analysis, this study compares prediction capabilities among AI models namely, Generalized Regression Neural Network (GRNN), Probabilistic Neural Network (PNN), Pattern Recognition Neural Network (PRNN) and Adaptive Neuro Fuzzy Inference System (ANFIS). The confusion matrices show that PNN performed better than other neural network models with 88.5% and 75.6% prediction accuracy in training and testing stages, respectively. Further comparison between PNN and ANFIS reveals that ANFIS outperforms PNN by prediction with 84.0% accuracy. Also, the R values of PNN and ANFIS prediction are 0.70788 and 0.79932, respectively. Whereas, the RMSE values for those models are 0.63607 and 0.50190, respectively. These also reveal the superiority of ANFIS than PNN model. Thus, ANFIS establishes itself as a best analytical tool for bus SQ estimation. According to the relative importance, the selected attributes are ranked with the AI models and public opinion. Connection weight method and Stepwise approach are followed to determine the relative importance of the attributes. According to all of the techniques, ‘Punctuality and Reliability’, ‘Seat Availability’, ‘Service Frequency’ and ‘Commuting Experience’ are found to be the most important SQ attributes. These outcomes of this study will convey an efficient way to the service providers, operators, policy makers and transportation authorities to improve the bus SQ in view of attracting more passengers. en_US
dc.language.iso en en_US
dc.publisher Department of Civil Engineering (CE) en_US
dc.subject Local transit-Mathematical model -- Dhaka City en_US
dc.title Application of neural network and fuzzy inference system for bus service quality prediction and attribute ranking in Dhaka city en_US
dc.type Thesis-MSc en_US
dc.contributor.id 1014042410 en_US
dc.identifier.accessionNumber 115056
dc.contributor.callno 388.4015110954922/KOK/2016 en_US


Files in this item

This item appears in the following Collection(s)

Show simple item record

Search BUET IR


Advanced Search

Browse

My Account