Abstract: This paper presented a method to predict the haze based on multiple linear regression analysis,whose sample is online update.First,the data of Beijing weather conditions are collected,including average temperature,humidity,wind and other meteorological data level and PM2.5,CO,NO₂,SO₂ and other atmospheric concentration data.Then,the main influencing factors are analyzed by scatter plot of these factors,and the main factors,which have a great effect on the haze,are selected as the haze forecast basis.Furthermore,PM2.5 prediction model is built by multiple linear regression method.The prediction results combined with meteorological factors are in the form of the criterion of haze judgment.Finally,this paper provided an actual example for haze prediction,which utilizes multiple linear regression to forecast the weather condition of Beijing after one day,three days and seven days.

Key words: Multiple linear regression,Main influencing factors analysis,Online update,PM2.5,Haze forecast