ChessAI Deep Learning Chess Move Predictor CNN

ChessAI-Deep-Learning-Chess-Move-Predictor-CNN

Este repositorio contiene un proyecto de Python que utiliza modelos de aprendizaje profundo para predecir los mejores movimientos de ajedrez dada una posición de tablero en el formato Fen (notación de Forsyth-Edwards). El proyecto utiliza redes neuronales convolucionales (CNN) para predecir el tipo de pieza, el alfa de Destination Square (columna) y el número (fila) de Destination Square para el siguiente movimiento.

Descripción general

The project includes the following components:

• Data preprocessing to convert FEN notation to a format suitable for deep learning.
• Three separate deep learning models for predicting piece type, alpha, and number of the destination square.
• Training and evaluation of these models using a dataset of chess game moves.
• A function to predict the best move for a given FEN board position.

Requisitos

To run this project, you'll need:

• Python 3.x
• TensorFlow and Keras
• Python Chess library ( python-chess )

You can install the required packages using pip with the following command:

 pip install chess tensorflow

Uso

1. Clonar el repositorio de su máquina local.
2. Train the deep learning models by running the respective training scripts.
3. After training, use the predict_best_move function to predict the best move for a given FEN board position.

 fen_board = 'r1bqkbnr/ppp2ppp/2n5/3Bp3/4P3/5Q2/PPPP1PPP/RNB1K1NR b KQkq - 0 1'
best_move = predict_best_move ( fen_board )
print ( "Best next move:" , best_move )

Model Files

The trained models (piece_model.h5, alpha_model.h5, and number_model.h5) can be made available on request at [email protected].

Feel free to explore and enhance this project for chess move prediction and analysis. Happy coding!