mxc666 / amuse Goto Github PK

AMUSE (Adaptive Mutation Strategy for Unsupervised Feature Selection) is a novel differential evolutionary algorithm designed to enhance the feature selection process for large datasets. It integrates an adaptive mixed mutation strategy for generating diverse mutation spaces and an assistance mechanism for selecting the optimal crossover probability ($CR$) during the crossover process.

• Adaptive Mixed Mutation Strategy: Dynamically generates mixed mutation spaces to enhance diversity and performance.
• Assistance Mechanism: Automatically selects the optimal crossover probability ($CR$) based on the standard deviation along the individual dimension.
• Improved Mutation Strategies: Incorporates 'local best' to reduce randomness and combines with existing mutation strategies to form a pool.
• Fitness Function: Integrates multiple indicators for effective unsupervised feature selection.

• Python: 3.8
• gensim: 0.13.4
• gym: 0.26.2
• gymnasium: 0.29.1
• numpy: 1.22.4
• scikit-learn: 1.3.2

The main_hypothyroid.py file serves as the main entry point for running AMUSE on the hypothyroid dataset.

1. Ensure all dependencies are installed.
2. Place the hypothyroid dataset in the appropriate directory.
3. Run main_hypothyroid.py using the command:

python main_hypothyroid.py

This script will execute the AMUSE algorithm on the hypothyroid dataset, showcasing its feature selection capabilities.