MATLAB Implementation of Cloud Classifier

This article discusses a classification program called the Cloud Classifier, which is implemented using MATLAB based on cloud model theory. The program demonstrates classification capabilities through testing on the Iris dataset. Key implementation aspects include cloud digital feature calculations (Ex, En, He) and forward cloud generator algorithms for handling uncertainty in data classification.

The Cloud Classifier's applicability extends beyond the Iris dataset to various other datasets, with its high accuracy attributed to cloud model theory's probabilistic characteristics. The implementation employs entropy-based feature weighting and hyper-entropy adjustments to enhance classification robustness. The classifier also offers strong interpretability through visual cloud drops distribution plots, helping researchers better understand dataset patterns.

Our implementation involves multiple technical stages: data normalization preprocessing using z-score standardization, feature extraction through principal component analysis (PCA), and model training with backward cloud generator algorithms. We experimented with different parameter configurations including cloud drops count (N=1000-5000) and entropy thresholds to optimize classification performance. The code structure includes modular functions for cloud transformation, similarity calculation, and classification decision making.

In summary, the Cloud Classifier proves to be an effective MATLAB-based classification tool that combines theoretical foundations with practical implementation. Our research provides executable code examples and parameter optimization methods to contribute to advancements in uncertainty-based classification techniques.