Our Methods

Method 1: Fart Gas Chromatography

Using our proprietary Fart-o-meter, we measured the gaseous emissions of 500 participants to create a comprehensive dataset.

Appendix: Fart Gas Chromatography Data

Method 2: Machine Learning Models

We trained a neural network on 10,000 fart samples to predict the most likely food culprits.

Model 1: Logistic Regression Model 2: Deep Fart Network

Method 3: Human Fart Judges

We employed 5 human judges to manually classify 2,000 farts as "good", "bad", or "mysterious".

Qualifiers Used by Human Judges