- Notifications
You must be signed in to change notification settings - Fork0
dcchang/ai-bme-final-project
Folders and files
| Name | Name | Last commit message | Last commit date | |
|---|---|---|---|---|
Repository files navigation
Artificial Intelligence in BME Final Project
Authors: David Chang and Mikey Komaiha
Date: April 2020
We evaluated different machine learning algorithms for predicting primary biliary cirrhosis patient outcomes.
This project used both unsupervised and supervised machine learning to predict patient survival time and status (alive or deceased) after a 10 year period.
Unsupervised learning used to explore data:
- Principal component analysis (PCA)
- K-means clustering
To predict patient survival time, we used the following models:
- Stepwise Regression [Performed best according to Pearson's correlation] --> Used subset of 12 noninvasive variables with log values substituted for albumin, bili, and protime
- Linear Regression
- Lasso
- Random Forest
To predict patient survival status, we used the following models:
- Logistic regression
- Random forest [Performed best according to Pearson's correlation]
- Support vector machine
Original data and reformatted data files contained indata
MATLAB live script (.mlx) and html output files contained inresults. We standardized all the data and ran the code again forfinalProjectZscore.html.
For full details, please viewFinal Report.pdf.
Original Research Paper:Prognosis in primary biliary cirrhosis: Model for decision making
Original Dataset:https://github.com/therneau/survival/blob/master/data/pbc.rda