An Intelligent Road Traffic Information System using Text Analysis in the Most Congested Roads in Metro Manila

• Published in: 2018 IEEE 10th International Conference on Humanoid, Nanotechnology, Information Technology,Communication and Control, Environment and Management (HNICEM) pp. 1 - 6
• Main Authors: Bondoc, Erika Ritzelle P., Caparas, Francis Percival M., Macias, John Eddie D., Naculangga, Vileser T., Estrada, Jheanel E.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.11.2018
• Subjects: Classification algorithms; Feature extraction; Latent Dirichlet Allocation; Machine learning algorithms; Named Entity Recognition; Real-time systems; real-time traffic detection; road traffic system; Roads; text mining techniques; Traffic congestion; Twitter; Twitter Streaming API
• DOI: 10.1109/HNICEM.2018.8666416

The increasing number of vehicles for every year in the Philippines produces a bigger percentage of causing severe traffic congestion in the three leading roads (Circumferential Road 5, EDSA, and Commonwealth Avenue) in Metro Manila. The researchers observed that most of the implemented technology used to detect traffic is by using a surveillance camera which requires labor and costly, while other web and mobile traffic information system used crowdsourcing and mostly owned by private. The researchers aim to develop a real-time road traffic information system by using the traffic-related tweets from MMDA (Metropolitan Manila Development Authority) official Twitter account to aid the current problem of providing sufficient road traffic information in the country. In this research, the traffic-related tweets from MMDA data are fetched using Twitter Streaming API, filtered by using Named Entity Recognition, preprocessed by applying tokenization, frequency counting and removal of unnecessary symbols, features were extracted by using Latent Dirichlet Allocation to identify the significant topic segments (time, day, lane of road, road direction, location and traffic mode) and Linear Regression was used for pattern recognition. Also, this study compares four machine learning algorithms (Naïve Bayes, Decision Tree, Random Tree, and k-Nearest Neighbor) for the proper identification of the most effective algorithm that will be used to for traffic mode classification. The experimental result shows that k-NN produced the best performance compared to other algorithms with 84.00% classification accuracy in anticipating the traffic congestion modes (Light, Light to Moderate, Moderate, Moderate to Heavy, and Heavy) in Metro Manila, Philippines.