Citation

Beura SK, Manusha VL, Chellapilla H, Bhuyan PK. Transp. Dev. Econ. 2018; 4(2): e11.

Copyright

(Copyright © 2018, Holtzbrinck Springer Nature Publishing Group)

DOI

10.1007/s40890-018-0066-0

PMID

unavailable

Abstract

This article proposes a bicycle level of service (BLOS) model for the assessment of urban roadway segments in mid-sized cities carrying heterogeneous traffic. The bicycling environments persisting on as many as 74 segments of four Indian cities are thoroughly analyzed. On-street bicyclists with varied demographics have rated these segments using a Likert scale ranging from '1' (excellent) to '6' (worst). The influences of various road attributes (geometric, traffic, and built-environmental) and bicyclists' characteristics (socio-demographic and travel characteristics) on the perceived ratings are assessed using Spearman's correlation analysis. Subsequently, eight significant variables are identified and used to develop a Levenberg-Marquardt neural network-based BLOS model. The most efficient but less complex model consisted of one hidden layer, three hidden neurons, and hyperbolic tangent activation function. This model produced very high values of correlation coefficient between the actual and predicted perceived ratings (i.e., 0.93 and 0.92 in the training and testing phases, respectively). The applications of Garson's algorithm and connection-weight approaches explored that the effective width of outermost lane has the highest influence on urban street BLOS. The BLOS criteria are classified into six categories A-F (representing excellent-worst) using the self-organizing map in artificial neural network cluster technique. It was observed that most of the studied segments are offering average to worst kind of services at their present-day conditions. Thus, the influencing variables should be largely prioritized in the planning process to achieve better service levels efficiently.

Language: en

Keywords

Artificial neural network; Bicycle level of service; Clustering; Heterogeneous traffic; Levenberg–Marquardt algorithm; Self-organizing map; Urban road segment