Decision Tree (DTs) classifier is most important and powerful solution of classification methods. One of the major problems in DTs is that they were built using crisp classes assigned to the training data. In the existing systems this drawback gets override with the concept of Emerging Pattern (EPs). Emerging pattern are those itemsets whose support in one class is significantly higher than their support in other classes. Hence DTs classifier are generalized along EPs so that they can take into account weighted classes assigned to the training data instances .The WDTs classifiers compared with other classifiers and proved that this methods have excellent noise tolerance and good performance. In the proposed system a new weighted decision trees classifiers is constructed using EPs and is compared with weighted Decision tree by applying Fuzzy feature ranking algorithm. Feature selection aims to reduce the dimensionality of patterns for classification by selecting the most informative instead of irrelevant and/or redundant features. In this paper, fuzzy feature clustering is proposed for grouping features based on their interdependence and selecting the best one from each cluster. Feature ranking is determined by means of different criterion functions. The accuracy and speed of both classifiers are evaluated, this comparative evaluation outsource which classifier has best performance.

">
 [PCST]
PCST Network

Public Communication of Science and Technology

 

An extended weighted classification technique using emerging patterns and feature ranking for breast cancer

Yamini C.   Anna University, Coimbatore

M. Punithavalli   SNS College of Engineering, Coimbatore

Decision Tree (DTs) classifier is most important and powerful solution of classification methods. One of the major problems in DTs is that they were built using crisp classes assigned to the training data. In the existing systems this drawback gets override with the concept of Emerging Pattern (EPs). Emerging pattern are those itemsets whose support in one class is significantly higher than their support in other classes. Hence DTs classifier are generalized along EPs so that they can take into account weighted classes assigned to the training data instances .The WDTs classifiers compared with other classifiers and proved that this methods have excellent noise tolerance and good performance. In the proposed system a new weighted decision trees classifiers is constructed using EPs and is compared with weighted Decision tree by applying Fuzzy feature ranking algorithm. Feature selection aims to reduce the dimensionality of patterns for classification by selecting the most informative instead of irrelevant and/or redundant features. In this paper, fuzzy feature clustering is proposed for grouping features based on their interdependence and selecting the best one from each cluster. Feature ranking is determined by means of different criterion functions. The accuracy and speed of both classifiers are evaluated, this comparative evaluation outsource which classifier has best performance.

[PDF 285.42 kB]Download the full paper (PDF 285.42 kB)

BACK TO TOP