Ordered Estimation of Missing Values for Propositional Learning

スポンサーリンク

概要

• 論文の詳細を見る

• When attempting to discover by extracting concepts embedded in data, it is not uncommon to find that information is missing from the dataset. The occurrence of such missing information can diminish the confidence on the concepts learned from the data set. This paper describes a new approach to filling missing values the missing values an attribute by using only the information contained in other attributes deter ignoring the class. Also an order is formulated for the construction of the decision trees used to determine the attributes, in an effort to reduce he computational cost. Experiments were conducted fro two different scenarios that differ in whether the test instances have missing values or not. Results on three datasets show that the new approach is successful in providing an input to the decision tree learning algorithm, which leads to final concepts with less error under different rates of random missing values. The approach is thought to be suitable for domains in which strong relations exist between the attributes, and for which it is worth trying to improve the accuracy in spite of the increased computational cost.

• 社団法人人工知能学会の論文
• 2000-01-01

著者

• Numao Masayuki

  Dept. Of Computer Science Graduate School Of Information Sciences And Engineering Tokyo Institute Of

• Numao Masayuki

  Dept. Of Computer Science Graduate School Of Information Sciences And Engineering Tokyo Institute Of

• Lobo Oscar

  Dept. Of Computer Science Graduate School Of Information Sciences And Engineering Tokyo Institute Of

• Ortega Lobo

  Dept. of Computer Science, Graduate School of information Sciences and Engineering, Tokyo Institute of Technology.

関連論文

• Suitable Domains for Using Ordered Attribute Trees to Impute Missing Values
• Ordered Estimation of Missing Values for Propositional Learning

スポンサーリンク