New feature released to show the accuracy of INSIGHT road traffic data

By Jerri Zhao

The most common questions we get asked about INSIGHT are, ‘Where do you collect your traffic data from?’, ‘What is your methodology for cleansing and processing the data?’, and ‘How can I be sure of the accuracy of your traffic data?’

INSIGHT’s new Confidence Indicator answers this last question.

The INSIGHT Confidence Indicator reveals how closely matched INSIGHT’s traffic volume data is to enhanced inductive loop detector data for each Local Government Area (LGA).

What is enhanced inductive loop detector data?

It is well known in the market that, despite its coverage, loop data is not consistent. Issues like outages, defective or downright broken sensors (plus other limitations on the accuracy of the technology due to heavy traffic) challenge the potential usability of this data. Before using any of it to validate our models, we go through an extensive exercise of cleaning and enhancing this data.

As you are likely aware, loop detectors function to record the number of vehicles travelling through a particular road lane. Many LGAs currently use this traffic data for planning and managing their safety projects and when implementing or making decisions about road and mobility infrastructure.

As this type of data is still commonly used by LGA, we have now included the difference between loop detector data and the INSIGHT traffic data in INSIGHT.

When you upgrade to INSIGHT Pro or Studio, in the ‘Streets Overview’ section, you will now see new fields in the ‘Volume by Road’ box (bottom right) or when you hover over a road on the map.

The ‘Error Margin’ (as shown above) is the difference between INSIGHT’s traffic volume data and inductive loop detectors. Almost every LGA in INSIGHT reveals an error margin of less than 20%.

The error margin increases or decreases by LGA. The lower the error margin, the closer the match is to the inductive loop detectors (for example, Campbell Parade in Bondi Beach shows a 14.5% error margin).

The error margins account for 80% of the roads that we publish. We guarantee that 80% of the roads that you see in INSIGHT are within the error margin stated.

As shown in the image above, the Middle Volume (28,950 trips) is the INSIGHT model’s traffic volume count for Campbell Parade, the Minimum Volume (24,752 trips) is INSIGHT’s traffic volume count minus the error margin, and the Maximum Volume (33,148 trips) is INSIGHT’s traffic volume count plus the error margin. This ultimately shows the range of traffic counts for that road, using the LGA’s error margin of 14.5%.

Here it is important to note that the reason we share the error margin with our INSIGHT Pro and Studio users is to maintain our level of transparency of data and methodology. We encourage our partners to make informed decisions about whether the INSIGHT data is fit-for-purpose.

In saying that, almost all error margins in INSIGHT are within a 20% range (with only a few LGA’s showing a worst case of 27% error margin). These error margins reveal that INSIGHT volume traffic data is sufficient and fit-for-purpose for most analytical and decision-making purposes.*

There’s now a quick and easy way to get the vehicle counts for a stretch of the road instead of conducting a time-consuming and costly project. And be confident of the results as well.

To ensure we publish the most accurate data covering all road classes (major and minor roads), our model training is optimised using the mean percentage error instead of r². This provides more stable behaviour when taking into account different types of roads. This method provides the error margin shown in INSIGHT; however, we can also report on r² if requested (as we have achieved r² above 0.95).*

If you have specific questions about our calculations or why r² is not our chosen method of calculation, click here to book a meeting with our INSIGHT expert Prashan.