An Integrated Platform for Mining Crowdsourced Data for Smart Traffic Prediction

Traffic prediction can help people make better travel plans by avoiding traffic jams, and also help the city to more proactively deploy emergency response vehicles. The continuous growth of social networks made possible the use of large amounts of data for traffic prediction. One of the biggest challenges in this regard is to acquire and process crowdsourced data to build effective models for traffic prediction. In this paper we propose a novel framework for processing crowdsourced data, with the goal of building effective traffic prediction models. We apply our solution to predict traffic related events in the busiest interstate in Colorado (USA), using Waze crowdsourced data. The events considered in the dataset are moderate jam, heavy jam, and stand still jam. In addition to traffic alerts crowdsourced data via Waze, we also use the traffic speed and weather data. The proposed solution proves to be effective and highly scalable, and the model's best accuracy on the test set is ~76%. This approach can be easily generalized in order to develop models that are able to provide effective traffic related predictions.