Traffic Congestion Prediction Using MATLAB and Machine Learning

A student-laptop project that predicts freeway congestion one hour ahead from a real traffic data set and studies short-term speed prediction and what-if scenarios on a simulated corridor. Hourly counts of the I-94 westbound station ATR 301 (UCI Metro Interstate Traffic Volume, 48,204 rows, 2012-2018) are cleaned and used to train baseline and machine-learning models; the forecast volume is turned into a volume-to-capacity ratio, an HCM level of service, a congestion level and a BPR travel-time index. A random forest forecasts the next hour with a mean absolute error of 143.63 veh/h (R2 0.9869) on the 2018 test year, 75.7 % lower than persistence. A cell transmission model of a 10 km three-lane corridor generates 120 days of loop-detector data; gradient boosting predicts detector speed 15 minutes ahead with 1.523 km/h error. A web application with five database options stores forecasts and simulations; the same study is written as MATLAB scripts.

Category: Civil Engineering / Transportation Engineering / Traffic Flow Theory / Machine Learning / MATLAB and Python

Civil Engineering Project C06 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.