Fig. 1
From: Singular spectrum-based matrix completion for time series recovery and prediction
Overview of the proposed sampling, recovery, and prediction scheme. On the left, the three images correspond to the spatial field at three different time instances, where the star symbols indicate the sensing nodes. During each sampling instance, red stars indicate sampling sensors while black stars indicate non-sampling sensors. The figure in the center depicts the incomplete measurement matrix where rows correspond to measurements from a specific sensor and columns to different sampling instances. The red square over the right part of the matrix highlights that in addition to missing value estimation, our system can generate a number of instances (columns) corresponding to future predictions. Individual sensor measurements are transformed to trajectory matrices that are introduced to the proposed SS-MC framework. The SS-MC algorithm produces completed trajectory matrices that can be joined to generate a fully completed (past and future) measurement matrix