Two-Phase Imputation with Regional-Gradient-Guided Bootstrapping Algorithm and Dynamics Time Warping for Incomplete Time Series Data

摘要

In this paper new algorithms with the combination between the Regional-Gradient-Guided Bootstrapping Algorithm and Dynamics Time Warping Technique for imputing incomplete time series data are proposed. The new measurement for curve similarity comparison by using the changing of slope of time series data are used. The main contribution of this paper is to propose new technique for imputing the fluctuate time series data. We compare our new method with Cubic interpolation, Multiple imputation, Windows Varies Similarity Measurement algorithms and Regional-Gradient-Guided Bootstrapping Algorithm. The experimental results showed that our new algorithms are outperform than these method.