Exploring and Analyzing Facebook as crowdsourcing platform for traffic updates using Selenium Support Vector Machine and Non-parametric LDA

Abstract

Traffic is a major problem in the Philippines. Facebook is one of the social media platforms that is commonly used by Filipinos. Machine learning is a field of computer science that allows computers to perform tasks like human beings. In this study, the proponents explored Facebook as a source of traffic updates and as a source of traffic information. In this paper, as a partial result, a machine learning model was created to classify Facebook posts as related to traffic. To gather Facebook posts, a total of 1000 respondents were asked for consent to scrape their public post using the username link and selenium. The Support vector machine model was trained with 3000 Facebook posts. The SVM model was only trained to 3 classes {Road accident, Road activities and Other}. The SVM model was evaluated using 10-cross fold validation. The result shows that the accuracy is 76% and the recall is 69%. To analyze the narrative of the corpus, the Hierarchical Dirichlet Process model was created with the log-likelihood of -4.06 with 10 topic models. The following are the narratives of the corpus: {Traffic Management, Immediate Emergency Response, Seeking help, Busses causes majority of accidents.}