Incremental Interval Type-2 Fuzzy Clustering of Data Streams using Single Pass Method

Area of Science:

• Data Science
• Artificial Intelligence
• Machine Learning

Background:

• Interval Type-2 Fuzzy C-Means (IT2FCM) is sensitive to initialization, often failing to find global optima.
• Optimizing IT2FCM with Ant Colony Optimization (IT2FCM-ACO) addresses initialization issues but struggles with large, evolving data streams.
• Existing methods are unsuitable for continuous, large-scale streaming data due to memory and processing limitations.

Purpose of the Study:

• To propose an incremental IT2FCM-ACO algorithm for efficient clustering of large data streams.
• To address the challenges of memory constraints and evolving data in streaming environments.
• To improve the scalability and performance of fuzzy clustering for big data.

Main Methods:

• An incremental approach is proposed, processing data in batches to determine cluster centers.
• Cluster centroids are updated iteratively with new incoming data points.
• Previous data points are released from memory to reduce time and space complexity.

Main Results:

• The incremental IT2FCM-ACO produces data partitions comparable to the standard IT2FCM-ACO.
• The proposed method demonstrates enhanced performance over other clustering algorithms on large, real-life datasets.
• Significant improvements in run time and excellent speed-ups were observed across all tested datasets.

Conclusions:

• The incremental IT2FCM-ACO effectively handles large data streams and evolving data patterns.
• The algorithm offers a scalable and efficient solution for fuzzy clustering in big data environments.
• The proposed method provides a robust alternative to existing clustering techniques for data streams.