Data mining for Breast Cancer Classification and Diagnosis Using Texture Features

Abstract

The data mining technique deals with the extraction of implicit knowledge with data relationship or other patterns not explicitly stored in the dataset. The main objective of this paper is to apply data mining on features extracted from mammograms to classify and detect the cancerous tissue. The data mining techniques are generally more suitable to larger databases than the one used for small dataset tests. Generally association rule algorithms adopt an iterative method to discovery frequent item set, which requires very large calculations and a complicated transaction process. Because of this, a modified association rule algorithm is proposed in this paper. Experimental results show that this method can quickly discover frequent item sets and effectively mine potential association rules. Texture features are most vulnerable features that include histogram intensity features and GLCM features which are extracted from mammogram images. A new approach of feature subset selection FSFCN is proposed which approximately reduces 50 to 60% of the features and the proposed association rule is used for classification. The most interesting one is that oscillating search algorithm which is used for feature selection provides the best optimal features and no where it is applied or used for GLCM feature selection from mammogram, Experiments have been taken from a data set of 322 images of MIAS of different types with the aim to improving the accuracy by generating minimum no. of rules to cover more patterns. The accuracy obtained by this method is approximately 97.82% which is highly encouraging.