Applying Expert Experience to Interpretable Fuzzy Classification System Using Genetic Algorithms

Accuracy and interpretability are two important objectives in the design of fuzzy classification system. In many real-world applications, expert experiences usually have good interpretability, but their accuracy is not always high. Applying expert experiences to fuzzy classification system can obtain better accuracy and preserve interpretability. In this paper, we present a method to translate expert experiences into fuzzy sets by similarity measure. Meanwhile reasonable experiences are integrated into a fuzzy genetic-based learning mechanism. Finally, experimental results with performance evaluation on benchmark classification problems demonstrate that the learning mechanism is able to achieve accurate performance for interpretable fuzzy classification systems.