Several researchers suggest that the Recommendation Systems (RSs) that are the "best" according to statistical metrics might not be the most satisfactory for the user. We explored this issue through an empirical study that involved 210 users and considered 7 RSs using different recommender algorithms on the same dataset. We measured user's perceived quality of each RS, and compared these results against measures of statistical quality of the considered algorithms as they have been assessed by past studies in the field, highlighting some interesting results.
Comparative evaluation of recommender system quality
P. Cremonesi;F. Garzotto;A. V. Papadopoulos;R. Turrin
2011
Abstract
Several researchers suggest that the Recommendation Systems (RSs) that are the "best" according to statistical metrics might not be the most satisfactory for the user. We explored this issue through an empirical study that involved 210 users and considered 7 RSs using different recommender algorithms on the same dataset. We measured user's perceived quality of each RS, and compared these results against measures of statistical quality of the considered algorithms as they have been assessed by past studies in the field, highlighting some interesting results.File in questo prodotto:
File | Dimensione | Formato | |
---|---|---|---|
Cremonesi-CHI2011-wip.pdf
Accesso riservato
Descrizione: Articolo principale
Dimensione
234.6 kB
Formato
Adobe PDF
|
234.6 kB | Adobe PDF | Visualizza/Apri Richiedi una copia |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.