作者: Lars Wohlrab , Johannes Fürnkranz
DOI: 10.1007/S10844-010-0121-8
关键词:
摘要: In this paper, we review possible strategies for handling missing values in separate-and-conquer rule learning algorithms, and compare them experimentally on a large number of datasets. particular through careful study with data controlled levels get additional insights the strategies' different biases w.r.t. attributes values. Somewhat surprisingly, strategy that implements strong bias against use values, exhibits best average performance 24 datasets from UCI repository.