A novel sampling-based visual topic models with computational intelligence for big social health data clustering.

Area of Science:

• Social media analytics
• Health informatics
• Data mining

Background:

• Twitter is a rich source of social health data.
• Topic models are used for health data clustering but require knowing the number of clusters.
• Existing visual techniques like MVCS-VAT are effective but computationally expensive for large datasets.

Purpose of the Study:

• To enhance the MVCS-VAT technique for efficient social health data clustering.
• To address the computational cost of assessing cluster tendency in big social health data.
• To improve the accuracy and efficiency of social health data cluster discovery.

Main Methods:

• Utilized a sampling technique to improve the multiviewpoint-based cosine similarity features VAT (MVCS-VAT).
• Developed and tested the proposed S-MVCS-VAT method on various health datasets.
• Evaluated the efficiency and accuracy of the S-MVCS-VAT against the original MVCS-VAT.

Main Results:

• The proposed S-MVCS-VAT method demonstrated improved accuracy of 5-10% compared to MVCS-VAT.
• S-MVCS-VAT proved to be faster and more memory-efficient for discovering social health data clusters.
• Experimental results confirmed the efficiency of the enhanced technique for big social health data.

Conclusions:

• The S-MVCS-VAT technique offers a more efficient and accurate approach to social health data clustering.
• Sampling techniques can effectively enhance existing methods for analyzing large-scale social health data from platforms like Twitter.
• This work provides a valuable improvement for health informatics research utilizing social media data.