Managing traffic congestion in developing and newly industrialised countries is challenging due to unreliable road infrastructure. In the South African context, electricity power cuts coupled with traffic light cable theft leads to severe traffic congestion. However, deploying traffic cops or pointsmen at critical intersections has been shown to dramatically reduce road traffic congestion. We develop a system that autonomously recommends the best distribution of human resources at critical intersections in order to improve road traffic flow. Our system combines traffic related Tweets with real-Time road traffic flow and incident data from TomTom® within an ensemble learning framework. Furthermore, a probabilistic ranking of traffic light intersections allows resources to be deployed to the most critical areas. Our system is developed and tested on real world data.