Deep Possibilistic C-means Clustering Algorithm on Medical Datasets.

Area of Science:

• Medical data analysis
• Machine learning
• Clustering algorithms

Background:

• Traditional clustering methods like Possibilistic C-Means (PCM) struggle with high-dimensional, large-scale medical datasets.
• The integration of IoT and big data in healthcare exacerbates computational complexity and feature extraction challenges in existing algorithms.
• Fuzzy C-Means (FCM) is susceptible to noise, while hard clustering provides rigid divisions, limiting their effectiveness in complex medical data.

Purpose of the Study:

• To propose a novel Deep Possibilistic C-Means (DPCM) clustering algorithm.
• To enhance the efficiency and accuracy of clustering for high-dimensional medical datasets.
• To overcome the limitations of traditional PCM and FCM in handling complex, large-scale medical data.

Main Methods:

• Integration of the Possibilistic C-Means (PCM) algorithm with an autoencoder deep network.
• Simultaneous optimization of deep neural networks and PCM clustering centers.
• Minimization of reconstruction loss via the autoencoder and utilization of soft affiliation for gradient descent in PCM.

Main Results:

• The proposed DPCM algorithm demonstrates improved clustering efficiency and accuracy on diverse medical datasets.
• Experimental results show superior performance compared to traditional clustering methods.
• The DPCM method effectively mitigates noise interference in medical data clustering.

Conclusions:

• Deep Possibilistic C-Means (DPCM) offers a robust solution for clustering complex, high-dimensional medical data.
• The synergistic combination of autoencoders and PCM enhances feature extraction and clustering quality.
• DPCM represents a significant advancement in medical data analysis, particularly in the era of big data and IoT.