https://github.com/fusi3/neural_networks_coursework

A comparison of MLPs and SVMs (using python) for the binary classification of heart disease. My final coursework at City, University of London for the neural networks course
https://github.com/fusi3/neural_networks_coursework

heart-disease mlp-classifier neural-networks python svm-classifier

Last synced: 28 days ago
JSON representation

A comparison of MLPs and SVMs (using python) for the binary classification of heart disease. My final coursework at City, University of London for the neural networks course

• Host: GitHub
• URL: https://github.com/fusi3/neural_networks_coursework
• Owner: fusi3
• Created: 2024-07-12T16:18:40.000Z (5 months ago)
• Default Branch: main
• Last Pushed: 2024-07-12T17:00:56.000Z (5 months ago)
• Last Synced: 2024-07-12T19:00:52.924Z (5 months ago)
• Topics: heart-disease, mlp-classifier, neural-networks, python, svm-classifier
• Language: Jupyter Notebook
• Homepage:
• Size: 2.41 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# neural_networks_coursework

I focus on critically evaluating the ability of two machine learning techniques – namely Multilayer Perceptrons (MLP) and Support Vector Machines (SVM) – to effectively classify whether a patient has heart disease or not. The problem is a binary classification problem with a relatively small number of features available. Four models are compared, two base models and two tuned models (one each for each technique). Base models are generated with simple starting inputs which were randomly selected while the tuned models were selected by tuning select hyperparameters through a grid search method. All models were crossvalidated with stratification and results were evaluated by comparing classification reports, confusion matrices and receiver operating curves (ROC). My Base SVM seems to be better suited for the task.

All the code and data files are uploaded to play around with. Install the libraries at the start of the notebooks for reproducability.