Outliers detection method for thermal sensation vote and its six influential factors in field survey

In the thermal comfort field survey,according to the characteristics of the six factors that affect human thermal comfort and the human subjective evaluation, this paper proposed a complete set of outlier detection methods innovatively. First, the six influential factors namely indoor temperature, relative humidity, globe temperature, air velocity, metabolic rate and clothing thermal resistance were cleaned. Then, based on the six factors and taking the standard effective temperature as the reference index, the d-based K-nearest Neighbor (KNN) classification combined with Gaussian distribution was adopted to deal with the anomalous value of TSV (thermal sensation vote). This method is an improvement on the existing research, which can detect the outliers in the data more comprehensively and objectively, so as to ensure the accuracy of subsequent data mining.