| Summary: | 博士 === 國立雲林科技大學 === 工業工程博士班 === 100 === Breast cancer is a malignant tumor that develops from cells of the breast. The incidence of breast cancer in women has increased significantly in recent years. Effectively methods that can accurately predict breast cancer are greatly needed and good prediction techniques can help to predict breast cancer more accurately. Thus, how to develop a more accurate model has become an important research topic. Therefore, the aim of this study is to propose a new method, cluster mixed with classification (CMC), that can be used for accurately diagnosing breast cancer to solve the problem in an attempt to predict results with better performance. In addition, the work tried to remove least important features to check whether it could help improve the results of breast cancer prediction, and to calculate the degree of importance ratings for features of breast cancer.
This work used non-parametric approach, Wilcoxon statistic test, to establish comparisons of the average AUC among the classifiers. From experiments 1 and 2, feature selection and ranking can thus provide clinicians with insight into their databases. The results of CMC are better than those of the other classifiers (SVM-SMO, KNN, and NB). Hence, it supports that CMC is an effective tool to predict breast cancer diagnosis. In addition, this work adopted ensemble methods aim to induce a collection of diverse predictors which are both accurate and complementary. The results demonstrate that ensemble classifiers are better than a single classifier. Generally, the proposed method method by this work can improve the performance of predictions and lead to further understanding of the disease manifestation.
|
|---|
