Activity index
The activity index is a Mapbox proprietary movement indicator that reflects the level of activity in the specified time span and geographic region. The index is calculated as follows:
- Raw data is aggregated by time span and geography, representing the total daily activity within each geographic unit.
- The de-identification process is applied. In this process, small random noise is applied to total counts and geographic areas with counts below a minimum threshold are dropped.
- A scaling factor is calculated by taking the 99.9th percentile of counts across all geographic unit and days in the baseline time span.
- Data is normalized by dividing the counts for each day and geographic unit.
- All normalized counts are rounded to the sixth decimal digit.
Baseline activity levels
Because the raw activity counts in later time span may be greater than activity counts during the baseline time span, the activity index has a range of 0 - ∞.
As an example, if the baseline time span is January 2020 and the 99.9th percentile count of total activity per grid tile in January 2020 was 1000 geolocation data points, then a grid tile that had 1500 data points on a day in March would have an activity index of 1.5 for that day.
Dynamic data sources
The Mapbox Movement data set is based on de-identified underlying mobile device activity that grows and changes every day. Any mobility data set is inherently skewed between highly populated regions (urban and metro areas) and sparsely populated regions (rural areas). So instead of providing "raw counts" of measurements, a custom normalization process is applied. This process measures and smooths out the impact of otherwise unpredictable changes in mobile device usage and calculates an activity index, which is more appropriate for comparison across time spans.