Decision Support modelling based on data mining for skid resistance

This project was a partnership between Queensland University of Technology (QUT) and the Queensland Department of Transport Main Roads (QTDMR) under the umbrella of CIEAM (CRC – Asset Management) with a focus on prevention of public roadway vehicle crashes by optimizing the road friction (skid resistance) of given road segments. The objective was to identify road segments from the whole road network that had a skid resistance deficit using data mining. In the identified roads the improvement of the road’s skid resistance would reduce the road’s crash rate.
This project charted new ground, because generally road crash studies focused on small homogeneous subsets of the road network because of the problem of providing quality data to support larger studies, and because of the complexity of describing the causes of road crash in a field of study still under investigation.

These problems were overcome in the following ways. Data mining was selected because of its ability to manage complexity, and to overcome the problem of the large proportion of road segments without skid resistance values, a novel method of applying data mining was developed and applied. The method created a road segment skid resistance/ crash profile that (1) used the data mining model trained on roads crash/skid resistance to plot the crash rate for the full range of skid resistance values, and (2) used the known crash rate to identify roads that could be improved, and provided an optimal skid resistance range for each road.

A prototype application was developed using the processes outlined to allow road asset managers to utilise the method. A sample profile of a 1 km road segment is shown in Figure 1, identifying the skid resistance curve, the actual 4 year crash count and skid resistance value of the road. An estimated optimal skid resistance value is estimated to reach the low crash threshold.