ACCIDENT BLACK SPOT IDENTIFICATION USING SOCIAL MEDIA DATA

Abstract

Traffic accidents are the most severe consequence of the expansion of road transport infrastructure. The road modifications are intended to be implemented in high-risk areas or accident blackspots where they will have the most impact; hence identifying these black spots is the first step in the safety management process. An Accident Blackspot can be theoretically defined as any location that has a higher number of crashes than other similar locations as a result of local risk factors. Traditional event detection methods are frequently hampered by a lack of sensor coverage, and reporting incidents to emergency response systems is time-consuming. Social media platforms such as Twitter can play an important role in giving way to gathering information on mobility and transportation services. Further, Sentiment analysis can be used to investigate the social media data and make valuable findings in transportation engineering difficulties. This report includes a collection of almost 65,000 tweets using Twitter APIs (Application Programming Interfaces), which are further classified into three different classes using Long Short Term Memory (LSTM) and Convolutional Neural Networks (CNNs) techniques, depending upon the textual information obtained from the tweets. The tweets are then geocoded by either using the location coordinates obtained from the Twitter API or the Named Entity Recognition and Part-of-Speech tagger hybrid model to identify the locations of the Accident-related tweets. Also, information about the type of vehicles involved in accidents is also obtained from the textual body of the tweets. These tweets are geocoded, and based on the information about the number of accidents in a particular area and accident severity, Accident Blackspots are identified.