A novel approach for recognition of control chart patterns: Type-2 fuzzy clustering optimized support vector machine

Area of Science:

• Industrial Engineering
• Machine Learning
• Data Mining

Background:

• Unnatural patterns in control charts indicate assignable causes of process variation.
• Pattern recognition is crucial for identifying and resolving process issues.

Purpose of the Study:

• To propose a multiclass support vector machine (SVM) based classifier for enhanced process variation analysis.
• To improve SVM effectiveness through integration with type-2 fuzzy c-means (T2FCM) clustering.

Main Methods:

• A fuzzy support vector machine classifier comprising fuzzy classifier, SVM, and optimization sub-networks was developed.
• The cuckoo optimization algorithm (COA) was employed for optimal hyper-parameter selection in SVM training.
• Type-2 fuzzy c-means (T2FCM) clustering was utilized to enhance SVM performance.

Main Results:

• The proposed system demonstrated very high recognition accuracy in simulations.
• The integration of T2FCM and COA significantly improved the SVM classifier's effectiveness.
• The fuzzy SVM architecture proved robust for pattern recognition in control charts.

Conclusions:

• The developed fuzzy SVM classifier effectively identifies process problems through pattern recognition.
• The proposed method offers a promising approach for real-time process monitoring and control.
• Optimized hyper-parameter selection via COA is vital for achieving high accuracy in SVM-based systems.