A Dimensionally Reduced Clustering Methodology for Heterogeneous Occupational Medicine Data Mining

Area of Science:

• Statistics
• Data Science
• Occupational Medicine

Background:

• Clustering identifies homogeneous groups within heterogeneous data.
• Principal Component Analysis (PCA) is widely used for dimensionality reduction but handles only quantitative data.
• Medical and biological data often comprise mixed quantitative and qualitative variables.

Purpose of the Study:

• To introduce clustering for multifactorial data analysis.
• To address the challenge of analyzing mixed-type data in occupational medicine.
• To develop strategies for simultaneous handling of quantitative and qualitative variables in clustering.

Main Methods:

• Utilized Principal Component Analysis (PCA) for dimensionality reduction.
• Developed projection strategies to integrate qualitative variables with quantitative ones.
• Applied clustering to PCA-regressed subspaces for mixed-type data analysis.

Main Results:

• Successfully integrated qualitative and quantitative data for clustering.
• Demonstrated the approach on an occupational medicine dataset of 813 individuals.
• Enabled pattern analysis in populations with heterogeneous measurements.

Conclusions:

• The proposed method effectively handles mixed-type data in multifactorial analysis.
• This approach advances clustering techniques for complex datasets, particularly in medical research.
• Offers a robust framework for pattern discovery in occupational health studies.