Breast Cancer Diagnosis Models Using PCA and Different Neural Network Architectures

Abstract

One of the most wide-spreading diseases among women is Breast Cancer. For this reason, a proper diagnosis is necessary for designating necessary treatment. Using the previous information about patients, diagnosis is being performed by various machine learning algorithms. As the data are getting bigger, it is becoming more necessary to extract the useful information from the huge pile of information. In this paper, we have used the Wisconsin diagnostic breast cancer dataset (WDBC) and SEER 2017 Breast Cancer Dataset. Then we have used Principal component analysis in order to extract useful features. After that, we have classified the reduced datasets using multi-layer perceptron (MLP) and convolution neural network (CNN). Then we have provided a comparative comparison of our model for both the reduced datasets. Our MLP model has achieved an accuracy of 99.1% on reduced WDBC dataset and 89.3% on SEER 2017 Breast Cancer dataset whereas CNN Model has achieved 96.4% on reduced WDBC dataset and 88.3% on SEER 2017 Breast Cancer Dataset.