Abstract: - in most developed nations, like our India, poorly managed roads are a reality in existence. For the well-being and prosperity of each nation a well-maintained road network is a must. Our focus is on the implementation of an efficient tracking system for road surfaces using one of the social media channels. Twitter is a social networking site that provides a vast number of facts daily, involving more than billion people. The most critical feature of Twitter is the potential for tweeters, which include incidents, scenarios, thoughts, views or even something completely different in real time. The social media tweet contains textual contents which enable users, including accidents and potholes, to view the lawsuits concerning different issues related to the transport of web sites. In order to take out and segregate tweets related to extraordinary problems, the keyword based approaches were formerly employed. These solutions are dependent solely on seed key phrases that can be provided manually. To address this problem, therefore, a special approach has been suggested, which uses the words2vec model to capture the semantic´ context through dense word embedding. However, the tweet separating method for the concept of equivalent semantic main phrases may still be subject to the question of pragmatic uncertainty. For this reason, the Word2Vec model is built for shaping tweets close to the semantinal ones. In addition, hotspots identical to each group have been established. These tasks help prevent accidents and can be used to identify danger spots prior to occurrence. Preventive movements can be alerted by the government and resources can be saved by preventive measures.Keywords— Road Traffic detection, Incident Detection, Social Media, Twitter, Machine Learning