A Study on Event Identification on Social Media Data

Abstract

This paper addresses the fact that there is no traffic tweet classification methods that try to identify the cause of the congestion. Our goal is to perform a multiclass classification of traffic-related tweets into traffic-congestion-cause-based groups. We perform various clustering techniques on our dataset, which we obtained from the Kolkata Traffic Police's Twitter handle. The clustering gives us the desired result of classifying the tweets into four broad categories based on the type of event causing the congestion.