On the Application of the Stability Methods to Time Series Data

The important problem of selecting the predictors in a high-dimensional case where number candidates is larger than sample size often solved by researchers from signal processing community using orthogonal matching pursuit algorithm or other greedy algorithms. In this work, we show how same can be applying methods based on concept stability. Even if it not new concept, stability less known community. We illustrate use presenting relatively family. As part presentation, conduct simulation study to investigate effect various parameters performance algorithm. Additionally, compare stability-based method with more eighty variants five different algorithms an experiment air pollution data. comparison demonstrates that leads promising results case.