Publication

Abstract : Death rate among women can be considerably brought down with regard to breast cancer if an early detection is viable. The prediction or detection of breast cancer in early stages is a complicated research problem. Using datamining techniques, it is not a difficult task to make it practical. The modern researches show that in most situations these techniques work better than common diagnostic methods. The basic aim of this work is to construct a data demonstrative model which can be used to: predict breast cancer survival even in the presence of missing values in the dataset that can reveal favorable information about the essential factors that determines the chances of survival, and also partition the patients with respect to their common peculiarities. Moreover, to find out a suitable filter-classifier combination. The Principal Component Analysis (PCA) and Decision Tree (J48) are chosen as filters. Further classification process is carried out on filtered dataset using the algorithms Logistic Model Tree (LMT), Random forest and Hoeffding Tree. Decision Tree (J48), were applied to choose the most efficient one. While implementing the classifiers, the dataset for which the feature selection is carried out using PCA gives better classification accuracies. The data mining tool WEKA provides a better platform for required experimental studies. A suitable filter - classifier pair is purposed for breast cancer prognosis by analyzing the results.