Short-term traffic prediction model for urban transportation using structure pattern and regression: an Indian context

Abstract

The paper presents a model for predicting short term traffic flow using structure pattern and regression. The relationship between traffic flow on the current section and upstream stations are used for predicting the short-term traffic flow and is exposed by the traffic flow structure pattern. The structure pattern of the traffic flow can be pinched from freeway toll data based on which, a new prediction model has been proposed. Locally weighted learning is used to predict the next instance’s traffic flow of current section by considering the traffic flow of current section and the next station entrance flow. This learning model deploys both linear and non-linear models to fit the nearby points and then applies those values to forecast the values of the query points. Experimental results illustrates that the short term prediction method based on structure pattern and regression is an effective approach for traffic flow prediction, this is applicable especially during the abnormal traffic states. The proposed system outperforms the conventional systems in terms of accuracy and proves to facilitate in conserving energy. The proposed work benefits the travelers by rerouting and saves fuel consumption.