Sketch Planning Level Shared Use Path User Volume Estimation using Machine Learning

Analyzing the volume of users in shared use paths (SUP) holds key importance for planners and policymakers. An in-depth understanding of the usage patterns of these paths is essential as it not only ensures the safety and convenience of users but also facilitates optimal path design and effective resource allocation. However, the advancement in technology has revealed a significant gap in comprehensive, accurate, and scalable techniques for estimating volumes among SUP users. In this study, data on SUP user volumes were collected from the Texas Bicycle and Pedestrian Count Exchange (BP|CX) and thoroughly compiled from multiple data sources. This enriched dataset serves as a valuable resource for predicting sketch-planning level demand along SUPs. The dataset was divided into subsets focusing on pedestrians and bicyclists and further categorized into temporal subsets, including daily counts, AM peak counts, and PM peak counts. For each of these scenarios, variable importance was assessed using gradient boost importance measures. Gradient boosting (GB), commonly known as multiple additive trees, is intended to enhance the accuracy of a learning algorithm by integrating multiple models with low error rates. Using the GB algorithm, the study identified the 50 most significant variables for each response variable (pedestrian or bicyclist AM peak, PM peak, or daily count). From these, 30 variables with the highest correlation to each response variable were selected through Pearson correlation coefficients. These 30 variables were then utilized in decision tree analyses to establish rules for sketch-planning level demand estimation models. The influential variables identified in this analysis include population data categorized by gender and income level, total household numbers, job accessibility, land use entropy, and per capita income, which have shown differences across bicycle and pedestrian models. Based on the availability of data for each of these variables, users can estimate sketch-planning level SUP pedestrian and bicyclist counts by day, the AM peak, and the PM peak using the decision tree rules generated for each scenario. This approach not only addresses the existing research gap but also provides essential insights for developing effective and practical models for SUP volume estimation at the sketch-planning level. The insights are essential for planners and policymakers to develop systematic and data-driven solutions by understanding how people use SUPs.