This paper develops a new mathematical-statistical approach to analyze a class of Flajolet-Martin algorithms (FMa), and provides analytical confidence intervals for the number F0 of distinct elements in a stream, based on Chernoff bounds. The class of FMa has reached a significant popularity in bigdata stream learning, and the attention of the literature has mainly been based on algorithmic aspects, basically complexity optimality, while the statistical analysis of these class of algorithms has been often faced heuristically. The analysis provided here shows deep connections with mathematical special functions and with extreme value theory. The latter connection may help in explaining heuristic considerations, while the first opens many numerical issues, faced at the end of the present paper. Finally, the algorithms are tested on an anonymized real data stream and MonteCarlo simulations are provided to support our analytical choice in this context.
Analytical confidence intervals for the number of different objects in data streams / G. Aletti. - In: BIG DATA RESEARCH. - ISSN 2214-580X. - 25(2021 Jul 15), pp. 100248.1-100248.11. [10.1016/j.bdr.2021.100248]
Analytical confidence intervals for the number of different objects in data streams
G. Aletti
Primo
2021
Abstract
This paper develops a new mathematical-statistical approach to analyze a class of Flajolet-Martin algorithms (FMa), and provides analytical confidence intervals for the number F0 of distinct elements in a stream, based on Chernoff bounds. The class of FMa has reached a significant popularity in bigdata stream learning, and the attention of the literature has mainly been based on algorithmic aspects, basically complexity optimality, while the statistical analysis of these class of algorithms has been often faced heuristically. The analysis provided here shows deep connections with mathematical special functions and with extreme value theory. The latter connection may help in explaining heuristic considerations, while the first opens many numerical issues, faced at the end of the present paper. Finally, the algorithms are tested on an anonymized real data stream and MonteCarlo simulations are provided to support our analytical choice in this context.File | Dimensione | Formato | |
---|---|---|---|
1909.11564.pdf
accesso aperto
Descrizione: Versione v2
Tipologia:
Post-print, accepted manuscript ecc. (versione accettata dall'editore)
Dimensione
812.59 kB
Formato
Adobe PDF
|
812.59 kB | Adobe PDF | Visualizza/Apri |
1-s2.0-S2214579621000654-main.pdf
accesso riservato
Tipologia:
Publisher's version/PDF
Dimensione
1.51 MB
Formato
Adobe PDF
|
1.51 MB | Adobe PDF | Visualizza/Apri Richiedi una copia |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.