Abstract
The purpose of the study was to develop a method for
making a self-adapting short-term
prediction model for the flow status. The model was based
on field-measured travel time data
and self-organising maps. The test site was Ring Road I
in the Helsinki metropolitan area.
The forecasts were based on the outcomes of previous
moments when the traffic situation
was similar to the present. The forecast was set to the
most common outcome in the cluster
of these similar samples. The model was allowed to work
online and its performance was
studied. The proportion of correct forecasts was
93.8-96.3% over the entire trial period and
80.9-82.3% in congested conditions for the model in
normal weather and road conditions.
The average daily change in the proportion of correct
forecasts was positive over the whole
trial period: +0.3-0.4%. Two naïve comparison models were
made. Both comparison models
performed considerably poorer than the self-adapting
model.
| Original language | English |
|---|---|
| Title of host publication | Proceedings of 12th WCTR |
| Subtitle of host publication | World Conference on Transportation Research |
| Place of Publication | Lisbon |
| Publication status | Published - 2010 |
| MoE publication type | A4 Article in a conference publication |
| Event | World Conference on Transportation Research, WCTR - Lissabon, Portugal Duration: 11 Jul 2010 → 15 Jul 2010 |
Conference
| Conference | World Conference on Transportation Research, WCTR |
|---|---|
| Country/Territory | Portugal |
| City | Lissabon |
| Period | 11/07/10 → 15/07/10 |
Keywords
- prediction
- traffic flow status
- self-organising map