Grey Combination Model Based On Wavelet Transform For Interval Prediction Of Infectious Disease

Interval prediction of number of infectious diseases is a prediction of the change range of number of infectious diseases over a period.To improve the prediction accuracy,the main work of this paper includes the construction of the interval prediction model and the application of the model to achieve interval prediction of number of infectious diseases.The main work is as follows:First,three new interval prediction models were proposed.Model 1: Vector regression combination model for interval-number time-series was constructed by combining vector autoregressive model for interval-number time-series and multiple linear regression model for interval-number time-series.Both internal factors and external system related factors of system characteristic factors were considered by this model.Model 2: The real parameters of GM(1,1)model were redefined as matrix,the prediction formula was solved according to Cramer's rule,the least square method was used to estimate the parameters,then the matrix multivariable grey interval prediction model was constructed.Model 3: Grey combination model based on wavelet transform was obtained by combination of wavelet transform,matrix multivariable grey interval prediction model and vector regression combination model for interval-number time-series.To analyze the approximate features and detailed features of interval-number time-series,wavelet transform was used to decompose interval-number time-series.The approximate interval-number timeseries after wavelet was predicted by the matrix multivariable grey interval prediction model,and the detailed interval-number time-series was predicted by the vector regression combination model for interval-number time-series.Second,after the completion of the construction of three interval prediction models,the three models were used to predict number of COVID-19 in the United States and number of TB in Xi 'an,China.The interval prediction model constructed in this paper was used to predict the daily number of COVID-19 in the United States.The interval-number time-series of daily number of COVID-19 in the United States showed an obvious exponential growth trend and fluctuated greatly in some moments;its accurate prediction can only be achieved with high prediction accuracy.By fitting three new models,GM(1,1)model and analyzing prediction,the grey combination model based on wavelet transform was more suitable interval prediction model in this study.The prediction results of this model showed that daily number will continue to increase,and the growth rate is fast.The monthly number of TB in Xi 'an of China was predicted by the interval prediction model constructed in this paper.The interval-number time-series of monthly number of TB in Xi 'an of China increased linearly from stable fluctuation and fluctuated greatly at some moments;only the interval prediction model with high prediction accuracy can accurately predict its trend characteristics and volatility characteristics.Using three new model,the multiple linear regression model for interval prediction,the GM(1,N)model with the system related factors;matrix multivariate grey prediction model,and the vector autoregressive model for interval-number,GM(1,1)model excluding system related factors to predict the monthly number.And contrasting the prediction,the grey combination model based on wavelet transform was the most optimal interval prediction model in the eight models.Forecast results show that the monthly number will continue to increase at a slow rate.