ASD Prediction Project

Overview

This repository contains our machine learning project for analyzing and predicting autism status using data from a questionnaire and some general demographic data.

This project includes:

• Requirements.txt - The python requirements needed to run the program
• ASD-Analysis.ipynb - The jupyter notebook including all of our work and data analysis
• 3 csv's for the input data:
  • test.csv - The main testing data
  • nonsynthetic.csv - The original dataset that the main testing data was generated from
  • test.csv - The test data, to be predicted to determine a final score

Project summary

The goal of this project was to predict whether individuals are ASD based on demographics and answers to a questionnaire, and to train/evaluate a variety of machine learning models including:

• Logistic regression
• Random forest
• XGBoost
• CatBoost
• Neural networks

Each model was tuned using Optuna for hyperparameter optimzation and evaluated using a bunch of metrics due to our heavy imbalance with an only 20% minority of the target class. We also explored feature importance to understand which factors were most important for the models.

Key observations:

At least according to our data set:

• 20% Of people are autistic
• Being white-european is heavily correlated with a high probility of an autism diagnosis
• People who didn't put down their ethnicity rarely have autism
• Being born with jaundice is somewhat correlated with an autism diagnosis
• Residing in the united states makes you more than 50% likely to have an autism diagnosis

Project Setup and Running Instructions

Note: Your python version must be 3.12, as TensorFlow doesn't support Python 3.13 yet.

Setting up the Virtual Environment

1. Create the virtual environment:

   • On Windows:

   python3.12 -m venv .venv

   • On Unix/MacOS:

   python3.12 -m venv .venv

2. Activate the virtual environment:

   • On Windows:

     .venv\Scripts\activate

   • On Unix/MacOS:

     source .venv/bin/activate

3. Install the required packages:

   pip install -r requirements.txt

Running Jupyter Notebook

1. Make sure your virtual environment is activated (you should see (.venv) in your terminal prompt)

   Note: You may be able to run without the virtual environment on windows, in which case you can skip steps 1 and 2 above.

2. Start Jupyter Notebook:

   jupyter notebook

3. Your default web browser should open automatically with the Jupyter interface. If it doesn't, you can copy and paste the URL that appears in your terminal.

4. Navigate to and open the ASD-Analysis.ipynb notebook to start working.

Deactivating the Virtual Environment

When you're done working, you can deactivate the virtual environment by running:

deactivate

Hackathon Project

The Team

Jake Malmrose - https://github.com/JakeMalmrose

Robbie Shirts - https://github.com/robertjshirts

Zachariah Hansen - https://github.com/ZachariahHansen

The Task

https://github.com/Aabo-Home/hackathon-2025-asd-analysis