Bayesian Inference with Medical Data

This project guides students through using Bayes' Rule, conditional probability, and Bayesian networks to reason about disease diagnosis from real-world-style medical data.

Project Overview

You'll explore how variables like symptoms, test results, and vaccination status affect medical diagnoses. The project includes:

• A synthetic medical dataset (.csv)
• A Jupyter notebook for guided analysis
• Exercises on joint/conditional probability and Bayes nets
• Visualizations of probabilistic dependencies

---

Setup (with Conda)

To install the required dependencies (including pgmpy, pygraphviz, and networkx), run:

conda env create -f environment.yml
conda activate cs3600-medical-analysis-env

Skills You'll Practice

• Estimating probabilities from real-world-style data
• Computing joint and conditional probability tables
• Applying Bayes’ Rule for probabilistic inference
• Building and interpreting Bayesian networks (Bayes nets)
• Constructing and using Conditional Probability Tables (CPDs)
• Identifying (in)dependencies between variables
• Making inferences with partial or uncertain information

---

Suggested Learning Path

1. Explore the dataset
   Understand the variables: diagnosis, symptoms, test results, vaccination status.

2. Compute joint probabilities
   Use pandas to create tables and uncover relationships between variables.

3. Use Bayes’ Rule to update beliefs
   Calculate the probability of a diagnosis given symptoms or test outcomes.

4. Build a Bayesian Network from the data
   Identify parent/child relationships and draw the graph structure.

5. Estimate CPDs
   Use grouped statistics to fill in tables like ( P(\text{Symptom} \mid \text{Diagnosis}) ).

6. Perform inference using the Bayes net
   Reason about unseen variables using observed evidence.

7. Reflect and extend
   Propose improvements, add variables, or try real-world scenarios.

---

Requirements

To run the notebook and analysis scripts, you need:

• Python ≥ 3.10
• Conda (Anaconda or Miniconda recommended)
• The following Python libraries (installed via environment.yml):
  • pandas
  • matplotlib
  • networkx
  • pgmpy
  • graphviz
  • pygraphviz
  • notebook