From the perspective of transportation planners the Seoul, metropolitan area is equipped with the world s best transit. ' Fare.Collection system. This system recognizes every passenger ', s origin transfers and destination, stops (or stations aswell.) As providing.Exact time stamps. However smart-card data, will not replace conventional household surveys until the trip purpose can. Be.Identified in a reliable manner. In this regard the present, study proposed a robust methodology to impute activities for. Smartcard.Data by using a continuous hidden Markov model (CHMM). The model uses unsupervised machine-learning technology.That requires no labeled data for training. When imputing, the purpose destination or mode, of GPS-based, location data. Many.Researchers have utilized various mathematical models that require a calibration procedure (Yang et al, 2010; Moiseeva. Et, al.2010; Allahviranloo and, Recker 2013a 2013b; Lu et al, 2013; Reumers et al, 2013; Liu et al, 2013), which computer scientists.Regard as supervised machine-learning technology. Furthermore prompted-recall surveys, have been a mainstream tool to obtain.Labeled data for calibrating and validating supervised imputation models (Feng, and Timmermans 2014; Giaimo et al. 2010;Greaves et al, 2010). Such surveys present the most probable activity to respondents and then ask them to check the correctness.And to fill in the details of the, true activity all of which is usually conducted using a portable electronic, device. The present study.Instead adopted an unsupervised model to recognize hidden activities behind a smart - Card Holder 's trip chain.The proposed unsupervisedmodel incorporated two critical tasks in imputing activities of smart-card data. That is clustering,,Activities was done simultaneously with deriving both membership probabilities for each cluster and transition probabilities.
การแปล กรุณารอสักครู่..