IDENTYFIKACJA UKRYTYCH KOMPONENTÓW W SZEREGACH CZASOWYCH METODAMI NIEUJEMNEJ FAKTORYZACJI MACIERZY
LATENT COMPONENT IDENTIFICATION IN TIME SERIES BY NONNEGATIVE MATRIX FACTORIZATION

Summary/Abstract: In this article, we present the use of non-negative matrix factorization methods to identify latent components contained in economic time series. A new non-negative matrix factorization algorithm based on Fermi-Dirac divergence is derived. Using the obtained latent components, we eliminate noise from the time series representing the prediction results. We test the entire concept in the problem of power system load prediction.

• Details
• Contents