Determination of the Receiver Operating Characteristics (ROC) Curve of the Logistic Regression Model Accuracy Using Some Breast Measurements in the Presence of Multicollinearity

Abstract

This study was aimed at determining the Receiver Operating Characteristics Curve of the Logistic Regression Model accuracy using some breast measurements in the presence of Multicollinearity to detect the presence or absence of tumorous cell of cancer patients. A secondary data from the Breast Cancer Wisconsin (Diagnostic) was used for analysis. The data was cleaned for outliers, recoded numerically and tested for multicollinearity. The ROC curve for the logistic regression model also revealed a high sensitivity and high specificity of the presence of tumorous cells in the patients with a percentage of 95% which is extremely high indicating that the logistic regression model is good in predicting better diagnosis of tumor cell of cancer patients accurately when combined with ROC.