Title :
Formal Analysis of Leave-One-Out Methods Based on Decremental Sampling Scheme
Author :
Tsumoto, Shusaku ; Hirano, Shoji
Author_Institution :
Dept. of Med. Inf., Shimane Univ., Izumo, Japan
Abstract :
This paper proposes a new framework for evaluation of cross-validation methods for rule induction based on statistical indices by using decremental sampling scheme, which is dual to incremental sampling scheme. The formal analysis of leave-one-out method for rule induction enables us to estimate error rate and other statistical indices, where the bias of error rate is represented by equations of given thresholds.
Keywords :
learning (artificial intelligence); sampling methods; cross-validation methods; decremental sampling scheme; error rate bias representation; error rate estimation; formal analysis; incremental sampling scheme; leave-one-out methods; rule induction; statistical indices; Accuracy; Educational institutions; Error analysis; Estimation; Probabilistic logic; Set theory; Training; accuracy; coverage; cross-validation; incremental sampling scheme; rule induction;
Conference_Titel :
Web Intelligence (WI) and Intelligent Agent Technologies (IAT), 2014 IEEE/WIC/ACM International Joint Conferences on
Conference_Location :
Warsaw
DOI :
10.1109/WI-IAT.2014.121
