Household travel data simulation: Application of spatial transferability of survey data

Due to the high cost, low response rate and time-consuming data processing, few Metropolitan Planning Organizations can afford collecting household travel survey data as frequently as needed. This study tested the feasibility of the spatial transferability of the National Household Travel Survey (NHTS) data by transferring the distributions from national level to a local area after updating.;Households in the national sample from the 2001 NHTS were clustered into 11 homogeneous groups based on the factors representing those 33 explanatory variables which include potential contextual factors as well as traditional socio-economic attributes. Using an Artificial Neural Network model, clustering process was reproduced for the add-on data sets of the NHTS.;To improve the accuracy of the transferability model, Classification & Regression Tree method was employed to further split each cluster to homogeneous subgroups. Another way of improving the cluster model employed in this study was local updating, with which a small local sample had the great potential to improve the travel statistics of the base model resulted from the NHTS national sample. The parameters of the best-fitted distributions were updated using an advanced Bayesian updating method with a small sample randomly extracted from the application area.;Meanwhile, a population synthesis was used to generate the entire population for the application area with the same 33 variables which were the inputs to the transferability model. These variables were later used to assign the cluster membership to the synthetic households. Then, the updated distributions of the travel attributes were used as the sources for Monte-Carlo simulation to generate synthetic household travel survey. By linking the generated travel estimates to the synthetic population, simulated household travel data for the application context were created.;Finally, the simulated travel data were compared against the add-on sample of the NHTS in the application area. The results of the comparisons showed very good fit between the means of the simulated and the validation data. Traditionally, transportation planners believed trip rates are easier to be transferred than any other travel statistics. However, this study showed that transferability of other statistics including trip length is also very promising.