CLASSIFICATION GROUPS FOR THE STAGE OF BREAST CANCER WITH STATISTICS METHOD

Abstract

This research is prediction of a model of possibility for detection of non- breast cancer and breast cancer and stages of breast cancers. Assessment of characteristics of abnormal growth of breast cancer cells such as Clump Thickness (X₁), Uniformity of Cell Size (X₂), Uniformity of Cell Shape (X₃), Marginal Adhesion (X₄), Single Epithelial Cell Size (X₅), Bare Nuclei (X₆), Bland Chromatin (X₇), Normal Nucleoli (X₈), and Mitoses (X₉) are independent variable. The dependent variable is the probability that the patient is detected at any stage of breast cancer or non-breast cancer based on the tumor cells with abnormal growth of breast cancer by using Ordinal Logistic Regression Model and Discriminant Analysis Model. In conclusions Ordinal Logistic Regression Model can use few variables in prediction of stages of breast cancer and Ordinal Logistic Regression Model has 55.50% classification higher than Discriminant Analysis Model has 54.10% classification. Ordinal Logistic Regression Model showed classification accuracy as following: non-breast cancer 73.60%, breast cancer stage 1- 4 as 5%, 43.6%, 61.4% and 76.4% respectively.  Nine predicted variables were used in Discriminant Analysis Model and classification accuracy was recorded as non-breast cancer 56%, breast cancer stage 1- 4 as 49.1%, 35.6%, 72.6% and 60% respectively.