IMPUTATION OF MISSING CLASSIFIED TRAFFIC DATA DURING WINTER SEASON

摘要

Highway agencies collect traffic data to calculate traffic parameters such as Annual Average Daily Traffic (AADT), Design Hourly Volume (DHV) and then to use as input in the planning, operation and management of their highway systems. The traffic data are usually collected through traffic monitoring programs. In particular, the Weigh-in-Motion (WIM) system is one of data collection systems to capture configuration patterns of vehicle travelling on the detection area. It is learned from literatures that traffic monitoring devices are prone to be in malfunctioning and, consequently, providing erroneous or missing traffic data due to the adverse weather conditions in which they operate. It is very critical for transportation agencies to be able to estimate classified missing traffic data in high accuracy level because the truck traffic plays a crucial role in developing pavement design and evaluation long term pavement performance. Several imputation methods have been cited in the literature but none of them have been designed to impute classified traffic data missed during severe winter weather conditions. To do this, winter weather model is structured and then calibrated to relate classified traffic volume variation to weather factors (snowfall and temperature) with traffic data collected from WIM stations located on highway network of Alberta, Canada and weather data collected from weather stations nearby WIM stations. Performance of the developed weather model is compared with a nonparametric regression method namely k-Nearest Neighbour (k-NN) method in terms of several error measures. It is concluded that winter weather models show better performance in terms of error measures than k-NN method while imputing the missing classified traffic data.