Breast Cancer Prediction
Machine Learning Project
Neural Network Model
Deep Learning models present an opportunity that we wanted to explore. We started with a flat model, moving further to a model with 2 hidden layers, 90 nodes (3-times the number of inputs), using a ReLu activation function and a Softmax output activation function. For the purposes of this project, we only visited the Adam Optimizer but it would be an interesting exercise to try out the AdaGrad optimizer here.
Neural Network Model with Full 30 Features
Neural Network Model with 7 BestFeature Selected Features
Neural Network Model with 7 Correlation Selected Features
Observations:
1. Model accuracy only declined slightly after reducing feature number from 30 to 7. The model with 7 correlation selected features suffered a little more.
2. Both 7 feature models were better at predicting Benign than predicting Malignant.