Cirrhosis Prediction
Notes and Code for implementation of Neural Networks
Project maintained by yashkathe Hosted on GitHub Pages — Theme by mattgraham
Cirrhosis Prediction
Table of Content
Overview
In this project, I have developed a predictive model to classify cirrhosis into four categories using a neural network built with PyTorch.
I’m using the Cirrhosis Prediction Dataset.
What is Cirrhosis?
Cirrhosis is a chronic liver disease characterized by scarring (fibrosis) of the
liver tissue, leading to impaired liver function.
It is commonly caused by long-term liver damage from factors like alcohol abuse,
hepatitis, and fatty liver disease.
Early detection and classification of cirrhosis stages are essential for timely
intervention.
Credits to the respective owner of the image
Files
Neural Network From Scratch
The files for this section include complete data preprocessing steps:
- Drop columns with excessive missing values
- Fill rows with missing values
- Data mapping
- Handle outliers
- Normalize the data
- Oversampling with SMOTE
Shallow Neural Network Using NumPy Only
=> Accuracy = 56%
Shallow Neural Network With Updated Hyperparameters
=> Accuracy = 66%
- Updated hyperparameters and the size of the hidden layer
- Broke down the backpropagation function into individual functions
Neural Network Using PyTorch
Neural Network With Gradient Descent
Neural Network With Stochastic Gradient Descent
Neural Network With Stochastic Gradient Descent and Dropout Layers
Random Forest Implementation
Random Forest With 275 Estimators
=> Accuracy = 44%
Notes on How Neural Network Works (From Scratch)
Extra
Open an issue if there are any improvements or errors in the code and notes.
⭐ Star the repository if you like the notes and approach.