Decision Support for Incident Management (DSIM)

Infrastructure, Transport and Logistics

"Making sense of 'big data' from multiple sources to optimise processes and procedures in control room environments
 

The NSW Department of Transport's Transport Management Centre (TMC) are using NICTA's advanced machine learning on big data issued from transport management for superior Decision Support for Incident Management in order to:

• Predict road clearance time and other relevant business factors early in incidents;
• Identify normal and abnormal patterns in historical incident resolution;
• Identify quantitative measures of task demand imposed on operators.

Benefits

Early prediction of incident characteristics will help prioritise communications with public and other agencies, provide decision support to the control operators during incident resolution, improve the design of existing software and standard operating procedures, and identify targeted operator training.

Scientific approach

Up to 7 years of historical data detailing all the steps taken in resolving each road traffic incident for the state of NSW will be analysed using a mixture of machine learning methods (including naïve Bayes, Classification Tree, Support Vector Machine and Bayesian non-parametric methods). Feature selection will represent an important aspect of the project due to the high complexity, and somewhat changing nature of the road network, combined with the variability in the response procedures. Machine learning methods will be adapted to the needs of spatio-temporal classification and prediction.

Previous work in the field

This project builds on NICTA's long-term relationship with the TMC, and relates to a number of other projects undertaken by our research team, including:

• DSIM 1: Measuring cognitive load of operators in traffic control rooms;
• STaRUI: Simplifying the user interfaces used by traffic control operators taking account of cognitive demand and user centered design to achieve improved incident clearance times;
• BushFire: Managing the cognitive load of operators handling multiple agencies in bushfire response centres;
• Ambulances: Decreasing demand on despatch operators, based on speech analysis.