Despite the widespread use, machine learning methods produce black box models. It is hard to understand how features influence the model prediction. We propose a novel explanation method that explains the predictions of any classifier by analyzing the prediction change obtained by omitting relevant subsets of attribute values. The local internal logic is captured by learning a local model in the neighborhood of the prediction to explain. The explanations provided by our method are effective in detecting associations among attributes and class label.

Deriving Local Internal Logic for Black Box Models / Pastor, Eliana. - ELETTRONICO. - 2161:(2018), pp. 1-4. ((Intervento presentato al convegno SEBD 2018 26th Italian Symposium on Advanced Database Systems tenutosi a Castellaneta Marina (Italy) nel June 24-27, 2018.

Deriving Local Internal Logic for Black Box Models

PASTOR, ELIANA
2018

Abstract

Despite the widespread use, machine learning methods produce black box models. It is hard to understand how features influence the model prediction. We propose a novel explanation method that explains the predictions of any classifier by analyzing the prediction change obtained by omitting relevant subsets of attribute values. The local internal logic is captured by learning a local model in the neighborhood of the prediction to explain. The explanations provided by our method are effective in detecting associations among attributes and class label.
File in questo prodotto:
File Dimensione Formato  
paper47.pdf

accesso aperto

Descrizione: Articolo principale
Tipologia: 2a Post-print versione editoriale / Version of Record
Licenza: PUBBLICO - Tutti i diritti riservati
Dimensione 136.46 kB
Formato Adobe PDF
136.46 kB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

Caricamento pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11583/2712649
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo