Traffic Congestion Detection from Twitter Using word2vec

Abstract

Event detection from social media is a highly important field of study. The main goal of the event detection is to detect events (e.g., traffic, disaster) automatically from social media like Twitter. In the area of supervised event detection from Twitter, word2vec model is not yet used by the researchers as a feature extraction technique. Since word2vec introduces a new approach on feature extraction from words and documents, in this work, we propose a near real-time traffic congestion detection system from the Twitter data stream using word2vec model as our feature extraction method. The system fetches the continuous stream of tweets from Twitter, preprocess and extract features from tweets using word2vec, classifies traffic congestion related tweets and notifies the occurrence of traffic congestion in a particular region. For the classification task, we compared the performance of Support Vector Machine (SVM), Logistic Regression (LR) and Naive Bayes (NB) algorithms. Our experimental results have shown that SVM outperforms the other two algorithms achieving an accuracy of 91.73%. Later on, the SVM has been selected to build the actual system.