.png){kind=icon}
.png){kind=icon}
Return
Understanding vehicle availability patterns using census data with spatially-hybrid machine learning models
.png){kind=icon}
PRE
AI
DOI：10.1016/j.cstp.2025.101641.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Predicting vehicle availability supports transportation planning by informing strategies to manage growing ownership and its effects on mobility and sustainability. This study applies spatially-hybrid machine learning models to explore vehicle availability patterns in Louisiana census tracts, a region characterized by a predominantly car-dependent culture and a mix of urban-rural dynamics, incorporating spatial factors such as geographic proximity and transit infrastructure for a comprehensive analysis. A custom spatial random forest (SpatialRF) model was used to select spatial predictors, which were then applied to train five ML models: Random Forest (RF), Artificial Neural Network (ANN), Support Vector Regressor (SVR), K-Nearest Neighbor (KNN), and Extreme Gradient Boosting (XGBoost). Key findings reveal that employment in the labor force, total working population, population aged 15-64 years, and median household income positively influence vehicle availability due to increased commuting needs and financial capacity, while smaller households (under 5 persons) and enhanced transit infrastructure reduce ownership by lowering transportation demand. The models exhibited strong performance, with R2 values ranging from 0.88 to 0.91 and low error metrics (RMSE, MAE, MSE), indicating high predictive accuracy and goodness of fit. Shapley Additive exPlanations (SHAP) were employed to interpret variable importance from the best-performing XGBoost model. This research highlights the limitations of traditional statistical models in capturing complex, non-linear relationships and demonstrates the potential of spatially-aware ML techniques to enhance transportation planning through more informed trip generation estimates, supporting better infrastructure and policy decisions.
Keywords:
Vehicle availability
Vehicle ownership
Census data
Journal
IF:
3.3
Papers:
181
Citations:
3.1K