The Preprocess of Time Series Data Based on Wavelet Transform

Outliers in time series data has a serious impact on the data analysis and use. Other methods to identify anomalies can´t identify and correct the first category and the second category of outlier at the same time. In order to solve this problem, this paper presents a new way to identify anomalies based on wavelet transform and identify outlier by the use of the wavelet transform modulus maxima , then pass the amendment of the outlier through inverse transform the wavelet transform coefficient. Evidence shows that this method is efficient in identifying and correcting the two types of outlier simultaneously.