Generation of the last forecasts: 16 January 2019 07:32:40 (UTC)
This model provides the prediction of the probability of traffic collisions on the roads of the City of Los Angeles. It is based on traffic reports provided by the LAPD. This database of vehicles collisions is combined with atmospheric, calendar and additional data to produce forecasts of the probability that a collision occurs in the coming hour for a 7-days period. The results are displayed on Google Maps from green for low probabilities to red for high probabilities.
This kind of models could be used in the future for the optimization of the global traffic of autonomous vehicles.
The map is initially displayed with the current prediction (at Los Angeles local time) of the probabilities of road collisions and is automatically refreshed every hour. Use the timeline bar (below the map) or the time control panel (on the map) to get a forecast at any other available time.
You can also change the region by moving on the map, change the scale or the background settings.
The minimum and maximum values of the probabilities are initially displayed on the map. You can display the forecast at any other place by directly clicking on the map. Use the button "Remove all markers" to delete all the displayed probabilities.
Real-time traffic conditions generated by Google Traffic are displayed only for predictions of the current time.
This project uses the "Traffic Collision Data from 2010 to Present" dataset (see below for details). These traffic collisions counts are combined with calendar and weather data from various models. Data science and machine learning techniques are then used to build predictive models and forecasts for future traffic collisions probabilities. This approach is transposable to the prediction of any weather-dependant variable.
All the data are provided by the Los Angeles Police Department and are available on this link under the Creative Commons CC0 Licence. This dataset is available for download from 2010 to present for the Los Angeles Area.
All the data generated by this experimental model are freely available without warranty of any kind.