Handwriting & Alzheimer’s disease: Diagnosis with machine learning (Econ 187)

by Jacob Titcomb

• This repository is the final project for Econ 187 (Machine Learning) at UCLA with Professor Randall Rojas, for Spring 2024.

• All work is my own.

• Python was the primary language used for this project.

• The data comes from a study by Cilia et al. (2022), entitled, "Diagnosing Alzheimer's disease from on-line handwriting: A novel dataset and performance benchmarking."

• All the work was done in the Jupyter Notebook econ187_proj3.ipynb. I wrote the final report in $\LaTeX$, with the output PDF econ187_proj3.pdf.

• This project serves to diagnose Alzheimer's disease based on kinetic data from written tasks. I picked this data because it seemed interesting.

Please enjoy!

Models

I fit the following models:

1. Nearest centroid (baseline)
2. Logistic regression
3. Decision tree
4. Random forest
5. Adaptive boosting (AdaBoost)
6. Extreme gradient boosting (XGBoost)
7. Support vector machine (SVM)
8. Neural network (MLP, for multilayer perceptron)
9. K-nearest neighbors
10. Linear discriminant analysis
11. Naïve bayes
12. Stacking
13. Voting