Document Type

Article

Abstract

A lazy learning method has relative advantages in comparison to eager learning method. However lazy learning has relative disadvantages also. Lazy learners are sensitive to irrelevant features. When there are irrelevant features, lazy learners have difficulty to compare cases. This is one of the most critical problems and the accuracy of reasoning can be degraded significantly. To overcome this restriction, feature weighting method for lazy learning have been studied. All the methods previously proposed tried to improve some parts of this generic process with different approaches. However, most of the existing researches were focused on global feature weighting. Therefore, we propose a new local method on e-business. The motivation to try local feature weighting method is that there are situations where locally varying weight vectors can help improving classifier performance by multimedia data model on e-business.

Download

Share

COinS