Early Detection of Breast Cancer Using Data Mining Algorithm Based on Historical Medical Records

Abstract

Nowadays cancer has become a common aspect and major source of death in the world. Most of the hospitals today don’t have a facility to detect breast cancer accurately. There is a possibility of cancer spreading more if proper treatment was not given for breast cancer for a long period. Therefore a proper diagnosis method must be developed earlier to detect the cancer and to avoid major death ratio. This paper aims at analyzing logistic regression using the Wisconsin Diagnostic Breast Cancer (WDBC) Dataset. The aim of this study was to construct a system weather a patient with a tumor is malignant or benign based on features and to optimize the learning algorithm. It focuses with feature selection/feature extraction methods and evaluates their performance to improve the accuracy. We find out the best feature characteristics and parameter values of our model with the help of machine learning logistic regression technique. By using normalization and fine-tuning of features, the proposed method increases the methods performance and finds the best model. The proposed approach obtained an accuracy of 89%, 90% and 92%.