An integrative machine learning framework for classifying SEER breast cancer.

Area of Science:

• Oncology
• Biomedical Informatics
• Machine Learning

Background:

• Breast cancer is a leading cause of mortality in women globally.
• Accurate prediction of patient survival is crucial for effective treatment strategies.
• Machine learning offers powerful tools for analyzing complex biomedical datasets.

Purpose of the Study:

• To classify the survival status (alive/deceased) of breast cancer patients.
• To develop a machine learning model using the Surveillance, Epidemiology, and End Results (SEER) dataset.
• To identify the most effective machine learning algorithm for breast cancer patient classification.

Main Methods:

• Data pre-processing and visualization of the SEER breast cancer dataset.
• A two-step feature selection process using Variance Threshold and Principal Component Analysis.
• Classification using supervised and ensemble learning algorithms: Ada Boosting, XG Boosting, Gradient Boosting, Naive Bayes, and Decision Tree.
• Performance evaluation using train-test split and k-fold cross-validation.

Main Results:

• The Decision Tree algorithm achieved the highest accuracy of 98% for predicting patient survival.
• This accuracy was consistent across both train-test split and k-fold cross-validation methods.
• The Decision Tree model demonstrated superior performance compared to other evaluated algorithms.

Conclusions:

• Machine learning, particularly the Decision Tree algorithm, provides a feasible and highly accurate approach for classifying breast cancer patient survival status.
• The feature selection method effectively reduced dimensionality while retaining important predictive information.
• The findings suggest that Decision Tree can be a valuable tool in breast cancer prognosis and patient management.