Distributed One-Class Support Vector Machine.

Area of Science:

• Machine Learning
• Data Science
• Big Data Analytics

Background:

• One-class classification is crucial for anomaly detection and outlier analysis.
• Traditional methods struggle with the scale and complexity of Big Data.
• ν-SVM offers a robust framework but requires adaptation for distributed environments.

Purpose of the Study:

• To develop a distributed one-class classification approach applicable to Big Data.
• To extend the ν-SVM method for parallel processing and global model generation.
• To enhance classification accuracy and computational efficiency for large datasets.

Main Methods:

• A novel mathematical formulation enabling separable optimization for distributed ν-SVM.
• Parallel processing of local data partitions on multiple processors.
• Inter-processor data exchange during learning for classifier specialization.
• Outlier-aware model generation ensuring robustness.

Main Results:

• The proposed method achieves high accuracy in decision region generation.
• Significant reduction in training time due to the distributed nature.
• Models are robust to outliers, unaffected by their position.
• Improved data fitting compared to existing classifiers.

Conclusions:

• The distributed ν-SVM approach is effective for Big Data one-class classification.
• The method offers a scalable and efficient solution for anomaly detection.
• Outlier-robustness and improved accuracy are key advantages.