Prediction of Metro Manila traffic from discrete time series

Abstract

We use correlation measurements to predict traffic at intersections of major roads in Metro Manila monitored by the Metro Manila Development Authority (MMDA) on their traffic monitoring site. Qualitative traffic data are recorded every 5 minutes from the site and are used to construct discrete time series. The Z measurement based on Pearson's r and mutual information score is used to pick out significantly correlated pairs of time series. A prediction method is applied to these pairs, and the success rate/characteristic curve is obtained for different time delays and using different lengths of historical traffic data.