. Introduction
The dynamics of jointly considering pollutants and traffic variables may have significant benefits for air quality ‘‘management’’
and conservation in urban areas, especially when integrated with intelligent transportation systems. Several simulation
models have been applied to link traffic flows to pollutant concentrations in urban areas. Moseholm et al. (1996) for
example considered volume, occupancy, speed and headway along with wind speed data to predict CO concentrations at
urban intersections and Nagendra and Khare (2006) trained static neural network (ANNs) to relate traffic composition to
NO2 concentration. Recently, Yang et al. (2008) considered traffic volumes and travel speeds to predict CO/CO2 levels an
Cai et al. (2009) predicted pollutants’ concentration based on factors such as traffic volume, composition, day of the week
and time of day, previous concentration levels, meteorological and geographical data.
Nevertheless, most efforts to incorporate real world traffic data to the process of pollutants prediction seem not to be systematic
or methodologically consistent with the short-term evolution of traffic flow; previous analyses on urban short-term
traffic time series data have shown evidence of non-stationary and nonlinear characteristics (Vlahogianni et al., 2006); this
statistical characteristics should be addressed, either explicitly or implicitly, during the modeling of the interaction between
pollutants and traffic flow. Moreover, the use of ANNs as more robust and accurate predictors of the anticipated pollutants
concentration, has been restricted to structural and learning formulations that are not consistent with the possible non-stationary
and nonlinear features of traffic and pollutants time-series. Almost all studies mentioned above concern multilayer
Perceptrons (MLPs) that are static in nature and contradict, at least conceptually, the initial consideration that both air pollutants
and traffic flow are, in essence, dynamically evolving.
. IntroductionThe dynamics of jointly considering pollutants and traffic variables may have significant benefits for air quality ‘‘management’’and conservation in urban areas, especially when integrated with intelligent transportation systems. Several simulationmodels have been applied to link traffic flows to pollutant concentrations in urban areas. Moseholm et al. (1996) forexample considered volume, occupancy, speed and headway along with wind speed data to predict CO concentrations aturban intersections and Nagendra and Khare (2006) trained static neural network (ANNs) to relate traffic composition toNO2 concentration. Recently, Yang et al. (2008) considered traffic volumes and travel speeds to predict CO/CO2 levels anCai et al. (2009) predicted pollutants’ concentration based on factors such as traffic volume, composition, day of the weekand time of day, previous concentration levels, meteorological and geographical data.Nevertheless, most efforts to incorporate real world traffic data to the process of pollutants prediction seem not to be systematicor methodologically consistent with the short-term evolution of traffic flow; previous analyses on urban short-termtraffic time series data have shown evidence of non-stationary and nonlinear characteristics (Vlahogianni et al., 2006); thisstatistical characteristics should be addressed, either explicitly or implicitly, during the modeling of the interaction betweenสารมลพิษและการจราจร นอกจากนี้ การใช้ ANNs เป็นทำนายอย่างสมบูรณ์ และถูกต้องของสารมลพิษคาดว่าจะความเข้มข้น ถูกจำกัดการก่อสร้างและสูตรที่ไม่สอดคล้องกับการได้เคลื่อนไหว ในการเรียนรู้และคุณสมบัติเชิงเส้นของการจราจรและมลพิษชุดเวลา การศึกษาเกือบทั้งหมดดังกล่าวข้างต้นหลายข้อกังวลPerceptrons (MLPs) ที่มีลักษณะคงที่ และขัดแย้ง กัน โดยหลักการน้อย การพิจารณาเบื้องต้นว่า ทั้งอากาศมลพิษและการจราจร ในสาระสำคัญ พัฒนาแบบไดนามิก
การแปล กรุณารอสักครู่..
