A Comparison Study of Missing Value Processing Methods in Time Series Data Mining

There are many methods for dealing with missing value on time series mining. The regression model is better than other methods when the variables of the data are correlative. This paper uses the method of mean interpolation and one variable linear regression, multivariate linear regression and iterative regression method of regression interpolation to deal with missing value of hydrological time series database and compares them under different Pearson correlation coefficient. The study shows that the one variable linear regression is simple and intuitive, and has a higher accuracy when data gaps exist variables that have a great relevance with missing variable values, and it shows that multivariate linear regression and multiple regression iteration have better results when the data doesn ´t.