Please use this identifier to cite or link to this item:
|Title:||Bayesian support vector regression for speed prediction with error bars||Authors:||Gopi Gaurav||Keywords:||DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence||Issue Date:||2014||Abstract:||Intelligent transportation systems (ITS) make use of modern technologies to improve and develop transportation systems. They help to improve urban mobility for commuters. In most metropolitan cities, traffic congestion is a serious issue and needs to be dealt with effectively. ITS can help to reduce traffic congestion by utilizing traffic prediction algorithms. The accuracy of predictions is key to the success of ITS. In order to have more robust performance, predicted values should be accompanied by measure of uncertainty associated with predicted traffic state. Machine Learning algorithms such as Support Vector Regress ion (SVR) perform traffic predictions with a high degree of accuracy. However, such methods do not provide any information regarding the uncertainty related to predicted traffic conditions. We can only calculate prediction error, once data from the field is obtained. To this end, we propose Bayesian Support Vector Regression (BSVR), which can provide error b.ar s along with the predicted information. This can helps ITS to overcome the problem associated with uncertainty in predictions. We apply BSVR to perform traffic speed prediction for multiple prediction horizons. We also employ BSVR to anticipate (detect) variations in prediction error. To analyze, detection performance of BS VR, we perform sensitivity and specificity analysis on prediction data. We discuss the performance of BSVR for expressways as well as general road segments.||URI:||http://hdl.handle.net/10356/65137||Fulltext Permission:||restricted||Fulltext Availability:||With Fulltext|
|Appears in Collections:||EEE Theses|
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.