A Novel Classification Algorithm Based on Incremental Semi-Supervised Support Vector Machine.

Area of Science:

• Computational Intelligence
• Machine Learning
• Pattern Recognition

Background:

• Traditional computational intelligence methods struggle with learning new concepts in dynamic environments due to inadequate data selection.
• Existing learning schemes lack mechanisms for dynamic data selection, hindering adaptation to environmental changes.

Purpose of the Study:

• To propose a novel classification algorithm inspired by human learning for adapting to changing environments.
• To develop an incremental semi-supervised learning system that addresses concept drift and improves classification accuracy.

Main Methods:

• Developed an incremental semi-supervised support vector machine (SVM) algorithm.
• Implemented a "soft-start" approach, a data selection mechanism, and a data cleaning mechanism based on prediction confidence and data distribution.
• Designed an algorithm that reduces computational complexity and handles new labeled samples.

Main Results:

• The proposed algorithm effectively utilizes unlabeled samples to enhance classifier knowledge and accuracy.
• Demonstrated a low rate of introducing incorrect semi-labeled samples.
• Achieved outstanding generalization performance and successfully overcame concept drift in changing environments.

Conclusions:

• The novel incremental semi-supervised SVM algorithm offers a robust solution for learning in dynamic environments.
• The method enhances classification accuracy and generalization by effectively managing data and adapting to concept drift.
• This approach provides a significant advancement in computational intelligence for adaptive learning systems.