Increasing generalization accuracy by using multivariate statistical method

摘要

Decision tree is one of the well-known approaches in the machine learning methodology. Researchers from various domains such as pattern recognization, data mining, statistics and machine learning are implementing decision trees for real time datasets. There is still a requirement for further efficiency and optimization. The problem of constructing decision tree with high generalization accuracy is still an active research area. Generating an efficient and optimized decision tree with multi-attribute data source is considered as one of the shortcomings. This paper emphasizes to propose a multivariate statistical method Discrete Wavelet Transform on multi-attribute data for reducing dimensionality and to transform traditional decision tree algorithm to form a new algorithmic model. The experimental results described that this method can not only increase generalization accuracy of the decision tree, but also improves the problems existing in pruning and to mine the better rule set without effecting the purpose of prediction accuracy altogether.