Elite Safety

Inspiration

Every year, there’s a huge monetary and personal loss to humanity, because of car crashes. So, when we heard about the hackathon, about the amount of data provided by VDOT and Smarter Roads, we as Computer Science students knew it right away, that we can contribute to this cause and decided to come up with a Machine Learning model to predict the number and the type of crashes.

What it does

In this design we used 789,914 car crash records to build a performance judging system to analyze the cause of each car crash and predict the number of car crashes on a particular road in future. To do this, we build a classifier and a regressor with machine learning techniques. The classifier was built on the crash records dataset using the Random Forest algorithm. When user, inputs the road features and real time conditions (like weather, time, etc..), 2000 estimators work to evaluate the possible cause of this crash. While the Classifier tell the user about the type of the predicted crash, the Regressor tells the user, the number of predicted car crashes in the future on a particular road. To build the regressor we cleaned and refined all the datasets and created a new dataset which records the crash and road statics of roads in Virginia. The regressor was built using linear regression algorithms. When user select the road and give the forecast weather in the next month, the regressor will return the expected crash numbers that will possible happen on this road in the near future.

Challenges Team ran into:

We collected a huge amount of data from the websites and combined several datasets, but still, we were, in need of more data to refine our model. Most of the data was raw data which need to be cleaned. To deal with these problem, we found several open database from the local government and ran statistical analysis on those data and create our own datasets.

What Team learned

We learned an enormous amount about the traffic conditions in the state of Virginia !

We learned a lot about data analysis and team working. We are excited about the future of autonomous driving and smarter roads and believe the future of AI and IT techniques in transportation.

What’s next?

We have built our classification model and regression model and made a web application for it. In the future, if we get more data, we can refine our model and predict more accurate analysis and predictions. And also, we will build a fully functioning application for car systems combined with voice control techniques.