PM2.5 prediction and early warning method and system based on nonlinear theory

The present invention discloses a PM2.5 prediction and early warning method and system based on the nonlinear theory. The method comprises a model training step and a model prediction step, and specifically comprises: dividing PM2.5 concentration time series data into two groups, and taking the two groups as a training time series data set and a test time series data set respectively; performing S-level wavelet decomposition and time-frequency analysis on the data of the training time series data set, expanding one-dimensional information into high-dimensional information, extracting implicitinformation of PM2.5 historical data, and obtaining a training time series index data set; constructing a predictive model; training the predictive model; performing MLRC-LSSVR model prediction on thetest time series data set, performing variance analysis on a model prediction result, and obtaining an upper critical value of the confidence interval as a final prediction result. The technical scheme of the present inventio