Class modelling by Soft Independent Modelling of Class Analogy: why, when, how? A tutorial

Area of Science:

• Chemometrics
• Machine Learning
• Data Analysis

Background:

• Soft Independent Modelling of Class Analogy (SIMCA) is a pattern recognition technique.
• Effective utilization of SIMCA requires understanding its principles and practical application.
• Existing resources may lack comprehensive guidance on SIMCA implementation and validation.

Purpose of the Study:

• To provide a comprehensive tutorial on Soft Independent Modelling of Class Analogy (SIMCA) for classification.
• To offer pragmatic guidelines for the correct and sensible use of SIMCA.
• To answer fundamental questions regarding the application, timing, and methodology of SIMCA.

Main Methods:

• Detailed explanation of the mathematical and statistical foundations of the SIMCA approach.
• Comparison of distinct SIMCA algorithm variants using two case studies.
• Development of a flowchart for optimizing SIMCA model parameters.

Main Results:

• Presentation of figures of merit and graphical tools for assessing SIMCA models.
• Inclusion of computational details and suggestions for SIMCA model validation.
• Introduction of a novel MATLAB toolbox for running and contrasting SIMCA versions.

Conclusions:

• The tutorial equips users with the knowledge to effectively employ SIMCA for classification tasks.
• The provided flowchart and validation strategies facilitate optimal SIMCA model performance.
• The accompanying MATLAB toolbox enhances the practical application and comparison of different SIMCA methodologies.