Crowd source based traffic analysis using machine learning algorithm

Abstract

One of the most detrimental effects to our economy currently is most certainly Traffic jam. Be it in a public vehicle or private, valuable work hours (around 3.2 million per day) is wasted everyday while waiting in the traffic. While this problem cannot be overcome without proper urban planning and traffic management, there are definitely ways of providing the commuters with an idea about how long they might be needing for their route. Often, this information might decide between rescheduling a meeting or missing it altogether. Keeping this in mind, the Dhaka Real Traffic project has been taken. It aims to provide travel time predictions based on machine learning of crowd sourced real commuting data. Data mining was done by means of a data collection app and also via Google form. The collected data was classed and trained by means of Python coding. From an initial choice between SVM, KNN & ANN, ANN was selected as the machine learning algorithm due to its lowest mean square errors among all three. Using Java and XML, the frontend Android App name Dhaka Real Traffic (DRT) was created with backend server learning. Due to machine learning, DRT will continue to upgrade its database to provide the most realistic travel time estimate.