摘要: |
According to Institute of Transportation Engineers, assessing the trip reduction claims from transportation demand management (TDM) programs is an issue for estimating future traffic volumes from trip generation data. To help assess those claims, a Worksite Trip Reduction Model and Manual was produced using existing data on programs, services and incentives contained in thousands of before and after worksite trip reduction plans. Models were built using linear regression and non-linear neural networks with the change in vehicle trip rate (VTR) as the dependent variable. No single variable selection technique, data handling method, or modeling approach yielded the best-fitting model for all three urban areas. The neural network model built on equally sampled data was the best generalized model based on three performance measures: the accuracy across the moderate range of change in VTR; the accuracy on full range of change in VTR and the R-square between the actual change in VTR and the predicted change in VTR. Worksite trip reduction plans explain a modest portion of the change in vehicle trip rates from one year to the next. The smaller datasets may have affected neural networks ability to identify correct non-linear relationship by overfitting the training data and reducing the neural network models power to generalize over unseen validation data. The aggregate nature of the data loses the ability to explain whether the change in mode behavior was influenced by the programs or changes in the workforce or other exogenous variables. Quality control issues with the provided datasets affected the model building process. Future research should examine organizational culture, management styles, the total expenses incurred by the employer, employee demographics and changes in the local economy. Finally, this Worksite Trip Reduction Model and Manual should be allowed to evolve like the ITEs Trip Generation Manual, which is in its sixth edition. Efforts to improve, maintain |