Thermal Comfort Voting(TCV)Prediction Based on Machine Learning
Taking a large shopping mall as the research site,the air state parameters,air quality pa-rameters,light environment parameters and sound envi-ronment parameters inside the mall were measured,and the personal parameters of the participants were meas-ured and surveyed.Machine learning algorithms(deci-sion tree,logistic regression,support vector machine and random forest)are used to establish a thermal comfort voting prediction model,and the impact of sin-gle input parameters and multiple input parameters on the prediction performance of the prediction model is analyzed.When the expected average evaluation model parameters are used as single input parameters,the pre-diction accuracy of the four prediction models is not i-deal,and the determination coefficient of the prediction results varies from 0.43 to 0.54,with significant fluc-tuations.The best performance is when the decision tree prediction model has air humidity as a single input parameter.Compared with other single input parame-ters,when air temperature and air relative humidity are used as single input parameters,the prediction perform-ance of the four prediction models is better.When air velocity and active metabolic rate are used as single in-put parameters,the prediction performance of the four prediction models is poor.Compared with the single input parameters,the prediction performance of the pre-diction model is not effectively improved when the ex-pected average evaluation model parameters are used as the multiple input parameters.After expanding the in-put parameters to all survey parameters,the prediction performance of all four prediction models is improved.The combination of input parameters,such as air state parameters,personal parameters and air quality param-eters can improve the prediction performance of the prediction model for thermal comfort voting.
万方数据
维普
