The purpose of this thesis is medical data classification using Python deep learning networks. Two types of neural networks were used in the study: convolutional neural network and recurrent neural network long short-term memory type. Both neural ne- tworks were implemented using the PyTorch library. The paper presents the results of experiments aimed at selecting appropriate network parameters in order to obtain the best classification accuracy. There were used four data sets: Breast Cancer, Breast Cancer Wisconsin (Diagnostic), Parkinsons, Heart Disease.
Datasets are provided from: https://archive.ics.uci.edu/ml/index.php
Other implememented functions:
- Adaptive Learning rate (https://uk.mathworks.com/help/deeplearning/ref/traingda.html)
- Cross-validation
CNN 1D Networks:
Paramaters test result for CNN:
Heart Disease Dataset:
Breast Cancer Dataset:
Parkinsons Dataset:
Breast Cancer Diagnostic Dataset:
