Data mining is a term used to describe the process of analyzing data from different perspectives and turning the results into useful information. The term is relatively new, but the concept of turning data into information is not. At some level, even aggregate summaries of measures such as RCT and ICT are examples of data mining. However, data mining can be much more detailed investigation into specifics of incidents or various incident types. Some of those interviewed in support of this guidance reported using data mining tools to examine the response timeline of particular types of incidents, such as fatalities, to determine how to improve the clearance times. Agencies that have utilized data mining techniques have typically done so to provide additional benefit to their incident management program, such as to:
- Simplify the aggregation of reports from multiple data sources and/or locations.
- Enhance the ability to compare two or more events in detail.
- Enhance the ability to peer into the data for underlying trends.
- Leverage results to use for other TIM activities, such as post-incident analysis of significant events.
It is beyond the scope of this guidance to examine the available data mining packages and make recommendations for a specific product line. The intent of the guidance in this regard is to communicate the opportunity that exists for understanding and improving TIM response with sophisticated data mining tools.
Data mining often goes hand-in-hand with visualization techniques. Visualization is simply using images to communicate a message, as was discussed in another section of this guidance. The figure below shows a data mining visualization from the University of Maryland Center for Advanced Transportation Technology (UMD CATT) that examines an individual incident in detail. The tool integrates information sources into a comprehensive timeline, including notifications, communications to the TMC, arrival and departure timeframes, and a depiction of which lanes were blocked during the entire timeframe of the incident. The live system is interactive using highlights, pop-ups, and a floating time pointer to specifically orient the user to every aspect of the response. The top portion of the figure shows a timeframe record of all the TMC communications related to the incident. The middle section shows the timeline of various responders and personnel associated with the incident response. The bottom section shows a record of the lane blocking characteristics of the incident, with red indicating that the lanes are blocked.
The figure below shows a heat map of the effects of the incident. The X-axis represents the timeline of the incident, while the Y-axis represents the distance upstream and downstream from the incident. The impact of this incident began at approximately 3:00 pm and extended more than 13 miles upstream from the incident location. This map integrates the messages posted on the roadway giving a comprehensive view of the traveler information response to the incident, in addition to the on-scene incident response timeline.
The figure below illustrates another type of data mining for TIM performance reporting. FAST recently began generating a “30-60-90” calculation for the Nevada DOT using the following categorization of incidents:
- An incident meets 30 minute criteria if it involves no injuries and it is removed from the travel lanes in 30 minutes or less.
- An incident meets 60 minute criteria if injuries are involved and it is removed from the travel lanes in 60 minutes or less.
- An incident meets 90 minute criteria if it involves a fatality and it is cleared in less than 90 minutes.
To aid in these calculations, FAST added a check box on the TMC incident screen for operators to indicate when an injury is involved (e.g., presence of an ambulance). In addition, FAST archives CCTV snapshot images taken during the incident timeframe (animation plays at 15-second intervals). Snapshots are taken of the incident location, as well as of adjacent roadway segments, such as ramps or arterial streets, that may also be impacted by the incident. By reviewing these snapshots, analysts can obtain additional details about incidents, as well as examine the impacts of the incidents on the roadways (which lanes are blocked/cleared and when). FAST has made use of these snapshot archives to help generate some of the 30-60-90 calculations for NDOT (e.g., to observe ambulance and fire responder activity to determine if an injury was involved). In the future, every time FAST provides an updated report to NDOT, the snapshots will be reviewed to provide the most accurate results as possible.
The take-away from any discussion on data mining is that there are multiple tools and techniques available, including commercially available products. The selection of features versus the trade-off of implementation and operations cost is a decision that needs to be made by each individual agency and/or by the TIM program partners. The model TIM performance measurement database discussed in a later section of this guidance illustrates the principles of data mining with realistic data and TIM performance scenarios.