Monte Carlo simulation of household travel survey data with Bayesian updating

Abstract

Several novel papers have demonstrated the feasibility of using Monte Carlo simulation to simulate characteristics that would be assembleed in a household travel observe - number of trips by means of purpose, mode of travel, trip longitudinal dimensions and time of departure - using sociodemographic data upon the household from a census, and distributions of travel characteristics from a national travel contemplate Although shown to be feasible as a mode the results tended to mirror still more of the nature of the source national travel data, and not as a great quantity [i]or[/i] amount of of the local characteristics as would be desired.

This paper describes research in which the distributions of travel characteristics were updated from a small local sample, using the course of Bayesian updating with subjective priors. The resulting distributions were ground to resemble more closely those of the local area for which the simulation was being undertaken. The paper describes the conduct of updating. It then summarises the proceeds obtained in updating the distributions for Adelaide in southern Australia, using original distributions from the US and local census data, together with a local household travel contemplate The paper also shows the issues of expanding the simulated data to the region. These be deriveds are compared with the expansion of actual overlook data for the region. It is lay the foundation of that the updated simulation gives a more accurate simulation follow for the region than simulation without updating.

This management offers a low-cost method for regions to unravel simulated household travel survey data that can be used for description of the region's travel and the exhibition of travel forecasting models, on the contrary at costs that are a fraction of those associated with normal household travel views The updating procedure requires a local sample of about 500 households, and the simulation can show as large a sample as may be desired for the region - up to the entire population - on the other hand at a cost of a scarcely any person hours of work.

INTRODUCTION

Household travel reviews (HTSs) are expensive to undertake, costing up to $350 through completed household in Australia, depending upon the data collection technique. owed to the large sample sizes (a minimum of 2500 to 3000 households) required to capture representativeness of existing household marks within both large and small metropolitan areas, the province of HTS are the large well-funded management organisations. Despite the absolute and relative take away froms HTSs are critical to travel-demand forecasts, used to estimate the impact of propos transport policies, and to defining regional travel patterns. HTS involve collecting socio-demographic data, with associated travel or activity data, for each of the observeed households which are then used to derive travel patterns for the take a view of region.

As with any take a view of method, public resistance to undertaking the scan task can result in high non-response rates and associated non-response errors. This may be amalgamateed if the survey is undertaken using a data collection course that has been shown to impact answer rates. Increasing personal mobility makes it more difficult to find individuals at their place of residence, where they can undertake the take a view of at a convenient time. The take away from of administering HTSs is increasing.

The research reported here exhibits a continuation of work undertaken by means of Greaves, Stopher, Bullock, and Kothuri upon the use of Monte Carlo simulation to create simulated HT data in a number of locations within the U A and in Adelaide, southern Australia (Greaves 1998, 2000; Greaves and Stopher 2000; Kothuri 2002; Stopher et al. 2003a,b). This research tread on the heels ofs on from the methodology used by dint of Kothuri (2002), who adapted the methodology unraveled by Greaves (1998), and is applied to the metropolitan area of Adelaide. The use of Adelaide go in the rear [i]or[/i] in the wake ofs the work undertaken by Stopher et al. (2003b) The process used in the earlier Adelaide research was based on drawing a sample of real households from the Australian Bureau of Statistics (ABS) Census data, for which there is known socio-demographic data, and then using socio-demographic travel pattern distributions obtained from a national HT to simulate the travel patterns within metropolitan Adelaide. The research described in this paper has the aim of repeating this task. However, the distributions used in this simulation are gained end Bayesian updating of the distributions from the national HT with local distributions obtained end a small sample. This is done to use the distributions inherent in the large national HT data plant while infusing these data with local characteristics from the local sample, thereby producing simulated data that better throw back the travel patterns in metropolitan Adelaide. A lock opener application of the simulation practice is the augmentation of existing HTS for travel demand forecasting of corridors and subregions. like an application would see a subset of households within a HT simulated, resulting in thousands of data points available to travel demand forecasters.