Szczegóły publikacji
Opis bibliograficzny
DEDSTA – an algorithm of density estimation for data streams with trend / Piotr KULCZYCKI, Tomasz RYBOTYCKI, Małgorzata Charytanowicz // Big Data Research ; ISSN 2214-5796 . — 2026 — vol. 45 art. no. 100605, s. 1–16. — Bibliogr. s. 15–16, Abstr. — Publikacja dostępna online od: 2026-05-14. — P. Kulczycki, T. Rybotycki - dod. afiliacja: Polish Academy of Sciences, Systems Research Institute
Autorzy (3)
- AGHKulczycki Piotr
- AGHRybotycki Tomasz
- Charytanowicz Małgorzata
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 169187 |
|---|---|
| Data dodania do BaDAP | 2026-07-31 |
| Tekst źródłowy | URL |
| DOI | 10.1016/j.bdr.2026.100605 |
| Rok publikacji | 2026 |
| Typ publikacji | artykuł w czasopiśmie |
| Otwarty dostęp |  |
| Czasopismo/seria | Big Data Research |
Abstract
This paper presents the innovative DEDSTA procedure, devoted to the estimation of distribution density for streaming data. It is particularly intended for cases when a trend, stable or with a varying intensity of changes, may appear in the investigated data stream. The characteristics of such a trend are identified within the designed method, as well as in consequence, special procedures are introduced and the parameters of the estimation algorithm are modified accordingly. Its concept is based on the methodology of kernel estimators. In cases when an essential trend appears, the algorithm employs components of statistical forecasting. Atypical elements (outliers) are also detected and the impact of those associated with new phenomena is strengthened and also the influence of diminishing is weakened. The proposed material provides an efficient method that is able to be used without additional extensive research and studies. Efficiency of the DEDSTA procedure was positively verified with the use of illustrative synthetic data as well as real data concerning meteorological measurements in Minneapolis, Rio de Janeiro, and Cracow, as example places with significantly different climates.
