This program attempts to model one's likelihood of having breast cancer with a neural network.
The dataset is used from scikit-learn: https://scikit-learn.org/stable/modules/generated/sklearn.datasets.load_breast_cancer.html
There are a total of 30 input parameters describing the tumor.
The model had:
- 3 hidden layers
- 300 in the first hidden layer with an activation of relu
- 60 in the second hidden layer with an activation of relu
- 30 in the third hidden layer with an activation of relu
- Output layer with 1 neuron with an activation of sigmoid
- BinaryCrossentropy loss computer
- Adam optimizer
- Learning rate of
5e-5 - 100 epochs
The results of the model was that it had a 94.7% accuracy with predicting a patient's likelihood of having breast cancer. Although the model predicts the likelihood of having breast cancer to a resonably high degree of accuracy, a 94.7% accuracy is still prone to error and not ready for clinical use.