MAPLSC: a novel multi-class classifier for medical diagnosis

Area of Science:

• Data Mining
• Bioinformatics
• Traditional Chinese Medicine (TCM)

Background:

• Clinical records are vital for expanding TCM knowledge and promoting its techniques.
• Multi-value classification is a significant data mining challenge, especially with multiple diagnostic syndromes in TCM clinical records.
• Existing methods struggle with imbalanced data distributions common in real-world datasets.

Purpose of the Study:

• To address the challenge of multi-value classification in TCM clinical records.
• To propose a novel and robust multi-class classifier designed for imbalanced data.
• To evaluate the performance of the proposed classifier against state-of-the-art methods.

Main Methods:

• Development of the Multiple Asymmetric Partial Least Squares Classifier (MAPLSC).
• Testing MAPLSC's robustness on imbalanced data distributions.
• Comparative analysis with seven other state-of-the-art classification methods.

Main Results:

• MAPLSC demonstrated significant improvements in multi-value classification tasks.
• The classifier showed robustness when dealing with imbalanced data.
• Superior performance was observed on both TCM clinical and public microarray datasets.

Conclusions:

• MAPLSC offers a promising solution for multi-value classification problems, particularly in the context of TCM.
• The proposed method enhances the analysis of complex clinical data with imbalanced distributions.
• MAPLSC represents a notable advancement in data mining for medical and biological applications.