Supervised machine learning models applied to disease diagnosis and prognosis

Area of Science:

• Computational biology and bioinformatics
• Medical data analysis and machine learning

Background:

• Accurate diagnosis and prognosis of cancer and heart disease are critical for patient outcomes.
• Machine learning (ML) offers powerful tools for analyzing complex medical datasets.

Purpose of the Study:

• To analyze and compare the diagnostic and prognostic capabilities of five ML algorithms on cancer and heart disease data.
• To identify key variables influencing cancer and heart disease.
• To evaluate the predictive accuracy of ML models using Receiver Operating Characteristic (ROC) curves.

Main Methods:

• Utilized five distinct Machine Learning (ML) algorithms for data analysis.
• Compared the predictive performance of each ML algorithm on breast cancer and heart disease datasets.
• Employed Receiver Operating Characteristic (ROC) curves to compute prediction accuracy.

Main Results:

• Identified significant variables contributing to cancer and heart disease.
• Random Forest (RF) demonstrated superior performance in breast cancer data analysis.
• Principal Component Regression (PCR) achieved the best performance for heart disease data analysis.

Conclusions:

• Machine learning algorithms can effectively aid in the diagnosis and prognosis of cancer and heart disease.
• Specific ML algorithms, such as RF for breast cancer and PCR for heart disease, show distinct advantages.
• Variable importance analysis is crucial for targeted prediction and understanding disease drivers.