Formulation of spatial temporal traffic information sequence and mining traffic sequence for prediction of traffic volume on highways

Abstract

Traffic management is an integral part of intelligent transport system (ITS). At present, the focus of the research community is on the innovation in technology-driven traffic management. Recent advancements in ITS have facilitated software enabled transportation infrastructure to the public. Internationally, transport industry strives to provide up to minute accurate travel information for hassle-free travel. Transportation researchers have widely acknowledged the dynamic and stochastic nature of traffic flow in short-term traffic forecasting. newlineA detailed literature survey on short-term traffic forecasting models and methods has been carried out. The application of deterministic models in traffic forecasting is presented abundantly in the literature. However, these models are limited in practical use due to complex mathematical structure. When such is the case, non-parametric regression method is an effective alternative. Machine learning techniques have also been found to be effective in traffic forecasting. However, traffic pattern analysis is essential when traffic condition is adverse and unprecedented. In this work, a sequential pattern mining approach is proposed in which traffic volume at the past is sequenced with travel time. Traffic information at successive time instance can be made available from past instances. Thus, traffic information representing the characteristics of traffic variable such as volume and travel time is used in the formulation of the proposed data-driven pattern mining approach. In urban transport, vehicles need to wait in a long queue during their travel due to which there is rise and fall in travel time. The delay in travel during rush hour needs specific consideration compared to non-rush hours of the day. Hence travel time in a journey is characterized by traffic volume on highways. newline newline