Chess Opening Classifier using Multi-Layer Perceptron Classification.
This project uses game data from January 2024 from lichess.org.
Here's roughly how it works:
1. Games are imported and the first few moves of each game are stored in a dataframe.
2. The dataframe is vectorized so that the moves are separated from the labeled opening.
3. The data is split between training data and test data using an 80/20 split
4. Uncommon openings are grouped together into an "Other" category since too few instances of these openings will make them unlearnable
5. The MLPClassifier is trained on the training data.
6. The trained model is tested using the test data
From the Lichess dataset I used 100,000 games, which leaves 20,000 for testing.
The Dataframe only cares about the first 5 moves of a game.
The Data uses ECO codes to classify openings, for example C51 is Evan's Gambit, and A40 is the queen's pawn.