Machine Learning for Classification of Traffic Congestion – New Traffic Management Perspective

With growing economic status, many people are buying vehicles which reflects their growing stature. Sadly, this is not backed by good traffic management systems which are increasingly leading to clogging of roads and unhealthy environment prevails leading to deterioration of health as well. Maps such as those provided by major players offset some of the problems faced by including traffic density in the path planning. But it still does not consider the weather conditions and emission levels in the area while suggesting a route. Also, the experience of past users of the route is not given any consideration of the route quality. This means that the estimation of traffic congestion goes beyond the accepted parameters of traffic density alone but is more intrinsically complicated than that. This also implies that there is no clear rule that can be written to classify the traffic congestion levels of the area. This means the only solution to estimate the traffic congestion levels of the region is to develop a robust machine learning model for the task of classifying the traffic congestion levels. By using this machine learning approach, we can use several feature parameters to map and model the traffic congestion levels to input and for which a rule-based approach fails as learning from past data is the only solution. This can result in better path planning and infrastructure making commute healthy, faster resulting in improved quality oflife.In this paper, we therefore apply a few classification models and determine its performance using the accuracy metric, training time and model stability. We therefore recommend the best fitting model from among the two models we have applied to this task of multi-class classification. We believe that this will transform the way transport is planned and pave way for a cleaner and healthier environment in the long run.