The data was taken from the Kaggle Competition

We have built a prototype app that uses a machine learning model that takes traffic, driver, road and weather data into account, predicts how dangerous a particular road segment is and alerts you.

The probability of a road segment being prone to accidents was computed with a deep neural network. The computed local 'danger levels' were then interpolated across the geography of a city network in Bayesian fashion, using Gaussian Process Regression.

The jupyter folder contains keras and data analysis notebooks. The resources contains the images of the analysis. The webapp contains the app's dashboard and a use case.