Research On Time-Activity Mode Evaluation Method Based On Automatic Monitoring Technology And Statistical Model

Objective:In the health impact and health risk assessment of air pollutants,more and more studies use the exposure assessment model to accurately assess the exposure level,and TAP information is an important input parameter of such exposure assessment model.Therefore,it is of great significance to obtain accurate TAP information for establishing exposure assessment model.In this study,students,retirees and employees with different commuting ways are selected as the respondents.GPS is used to collect information(x)related to TAP,and GPS track display is used to modify the TAD questionnaire information to obtain accurate TAP information(y)of the respondents,and a TAS classification algorithm based on RF model is constructed,and then a TAP evaluation method suitable for individual exposure assessment of air pollutants is established.Method:1.In the urban area of Beijing,the respondents are recruited by means of volunteer recruitment.The subjects cover three groups with different TAP primary school students(n=23),employees(n=28)and retirees(n=33).At the same time,considering that there may be differences in TAP between four seasons,the retirees and employees are investigated once in summer and once in winter,with each investigation covering at least one working day and one holiday.2.The basic information of the respondents is obtained through the questionnaire survey,including age,education level,family and workplace,ownership of vehicles(private car,electric bike and bicycle).3.Collect the TAS information of the respondents through the TAD,including the time,location,activity,travel mode and means of transportation(walk,private car/taxi,subway,bus,electric vehicle,bicycle,etc.)of each travel start and end point.4.Collect the location coordinates(longitude and latitude and altitude),moving speed and moving direction of the respondents by carrying GPS,and extract the motion parameters or position parameters information weighted by different time as candidate prediction variables(x),mainly including distance from home address,distance from workplace or school,distance from road,Different time weighted(1-15 min)average moving speed,maximum or minimum speed and variation of moving direction.5.Through the combination of the dynamic display of GPS track on the map(Google Earth and leaflet map)and the TAD questionnaire information,the accurate determination of the TAS of the respondents is realized;the TAS is classified according to the movement characteristic index under different TAS and the difference of air pollution level in the micro environment.6.Taking the accurate TAS(time resolution is lmin)as the dependent variable,the moving speed,direction and distance extracted from GPS data as the independent variables,and using the RF algorithm to establish the identification model for the TAS of the respondents;using the 10 fold cross validation method to evaluate the prediction effect of the model.Result:1.Children,employees and retirees show different characteristics of time activities The elderly people spend more than 86%of the whole day in the residential room,91%time in winter which is higher than 87%in summer.The average time spent in other activities is less than 5%,and TAP is similar on the working day and on weekend;The employees spend more than 5%of the whole day in the residential room on the working day 56%(summer)-62%(winter),secondly on workplace and transportation;on weekends 81%(summer)-88%(winter),secondly public indoor;The primary school students spend 65%time in family,21%in school and 5%outdoor during a week2.In the model training stage,the overall accuracy of the prediction for three groups of people's different activities is 93.2%-97.7%;in the model verification stage,the overall accuracy of the prediction is 77.5%-88.7%,which has relatively high.3.The prediction accuracy of the model for different TAS is different.The TAS with the highest prediction accuracy are in the family and workplace(or school),secondly in public places and on motor vehicle travel,and the prediction accuracy in outdoor or on non-motor vehicle travel is the lowest4.The main prediction variables of the model include distance from road,home address and workplace(or school),time weighted average speed,variability of moving direction and ownership of family vehiclesConclusion:The results of this study show that the TAP assessment method based on GPS automatic monitoring technology and RF algorithm can realize the time state discrimination of different populations,and has high accuracy,and can meet the basic requirements of air pollutant exposure assessment.