search query: @supervisor Heljanko, Keijo / total: 36
reference: 5 / 36
| Author: | Rancati, Pietro |
| Title: | Vehicle Travel Time Prediction based on Car Fleet Data and Additional Data Sources |
| Publication type: | Master's thesis |
| Publication year: | 2016 |
| Pages: | 63 s. + liitt. 2 Language: eng |
| Department/School: | Perustieteiden korkeakoulu |
| Main subject: | Cloud Computing and Services (SCI3081) |
| Supervisor: | Heljanko, Keijo |
| Instructor: | Vehkomäki, Tuomo |
| Electronic version URL: | http://urn.fi/URN:NBN:fi:aalto-201608263129 |
| Location: | P1 Ark Aalto 4402 | Archive |
| Keywords: | intelligent transportation systems machine learning algorithms travel time prediction traffic information |
| Abstract (eng): | Travel time is a fundamental measure in transportation. Accurate travel-time prediction also is crucial to the development of intelligent transportation systems and advanced traveller information systems such as traffic monitoring, finding driving directions, ride sharing and taxi dispatching. The travel time of a vehicle is influenced by several factors that can bring to large difference between the standard travel time and the travel time in particular situations. These factors can depend from the vehicle itself, for example driver behaviour, and from external factors, such as traffic jams and weather. In this thesis, we realized a model that is able to generate accurate travel time tables for road segments depending on the day of the week and the time of the day.Initially GPS data from a car fleet is processed through a map matching algorithm in order to obtain the travel time records done by drivers on road segments of the city. Subsequently, a street categorization algorithm is realized to determine the average time required on each road segment of the city. In order to cope with the different amount of data depending on the road segment, a machine learning model and interpolation techniques are applied. The travel time tables generated are processed by two machine learning models in order to consider different weather conditions and driver's behaviors that can influence the predicted travel time. Compared to other travel time tables provided by commercial competitors, our results show that the model created, with the inclusion of factors that can influence the predicted time, can significantly reduce both relative mean errors and root-mean-squared errors of predicted travel times. We demonstrate the feasibility of applying our model in travel-time prediction of fleet management system and prove that it performs well for traffic data analysis. |
| ED: | 2016-09-04 |
INSSI record number: 54340
+ add basket
INSSI