Online Knowledge-Based Model for Big Data Topic Extraction.

Area of Science:

• Machine Learning
• Artificial Intelligence
• Data Science

Background:

• Lifelong machine learning (LML) models continuously learn from experience without user intervention.
• Traditional LML models struggle with high data dependency, resource consumption, and lack support for streaming data.
• Scaling LML models for big data analysis presents significant challenges.

Purpose of the Study:

• To propose an online lifelong machine learning (OAMC) model designed for streaming data environments.
• To address the limitations of existing LML models, including high data dependency and resource inefficiency.
• To improve the learning pattern of LML models for data arriving in sequential pieces.

Main Methods:

• Engineering the knowledge-base within the LML model.
• Introducing novel knowledge features to enhance the learning process.
• Developing an online adaptation mechanism for continuous data streams.

Main Results:

• The OAMC model demonstrates improved accuracy in topic coherence by 7% for streaming data.
• Processing costs are reduced by half compared to traditional LML approaches.
• Reduced data dependency and enhanced learning patterns for piece-wise data arrival.

Conclusions:

• The proposed OAMC model effectively supports streaming data in lifelong machine learning.
• OAMC offers a more efficient and accurate solution for big data analysis with continuous data streams.
• The advancements in knowledge-base engineering and feature introduction are key to OAMC's performance.