Gap-Filling in Time Series by Multivariate Singular Spectrum Analysis

This paper, introduce a new algorithm using Multivariate Singular Spectrum Analysis (MSSA) for gap filling in univariate time series. In this algorithm, the data before missing values and the data after (in the reverse order) missing values treated as two separate time series. Then using MSSA forecasting, two estimations of the missing values will be provided. Finally, using bootstrap we combine these estimations and produce a unique estimation for missing values.