Dealing with Multicollinearity Problem in Analysis of Side Friction Characteristics Under Urban Heterogeneous Traffic Conditions

Disturbance to the moving traffic, such as on-street parking and bus stop on carriageway normally called as side friction in the literature, is one of the major problems that impact the vehicular speed on urban arterials in developing countries like India. Regression model to study the mode-wise vehicular speeds using side friction and other influencing variables faces the problem of multicollinearity as there is a strong correlation between the independent variables which results in high R-2 and high p values. Instead of dropping the highly correlated variables which is the conventional practice to deal with multicollinearity, we proposed an approach of taking a linear combination and ratio of independent variables. To test the applicability of the proposed approach, traffic data from five locations in Hyderabad and Vellore in India having a wide variety of geometric and traffic characteristics with one or more side frictions were collected and regression models were built. It was found that the proposed approach of taking linear combination and ratio worked very well as the models exhibited a high R-2, low p values and low variance inflation factor. The regression models also forecast speeds with good and reasonable accuracy. For better understanding of the impact of side friction, the present study proposed an index called side friction index and it was found that the number of side friction elements and their impact are directly proportional. Also, the impact is not uniform across the different vehicle types. Though the present study focussed on side friction problem, the approach that we proposed in this paper to deal with multicollinearity can be applied in other fields as well.

Automated summary

In developing countries like India, side friction in traffic, such as on-street parking and bus stops on carriageway, is a major issue impacting vehicular speed. The proposed approach of using linear combinations and ratios of independent variables effectively addresses multicollinearity issues and provides accurate speed predictions.