Comparing supervised and semi-supervised Machine Learning Models on Diagnosing Breast Cancer

Area of Science:

• Computational biology
• Medical informatics
• Machine learning in healthcare

Background:

• Breast cancer is the most prevalent cancer among US women.
• It is also the second leading cause of cancer-related mortality in women.

Purpose of the Study:

• To evaluate and compare the performance of supervised and semi-supervised machine learning algorithms for breast cancer prediction.
• To assess the accuracy and effectiveness of various classification algorithms in diagnosing breast cancer.

Main Methods:

• Employed nine supervised (SL) and semi-supervised learning (SSL) algorithms, including Logistic Regression, Naive Bayes, SVM, Decision Tree, Random Forest, XGBoost, Gradient Boosting, and KNN.
• Utilized the Wisconsin Diagnosis Cancer dataset for model training and testing.
• Applied K-fold cross-validation and hyperparameter optimization to ensure model robustness.
• Evaluated models using accuracy, precision, recall, F1-score, and ROC curves.

Main Results:

• Both SL and SSL models demonstrated strong performance.
• SSL algorithms achieved high accuracy ranging from 90% to 98%, utilizing only half the training data.
• The KNN model (SL) and logistic regression (SSL) achieved the highest accuracy of 98%.

Conclusions:

• SSL algorithms exhibit accuracies comparable to SL algorithms, with overall model accuracies between 91% and 98%.
• SSL presents a promising and competitive alternative for breast cancer diagnosis.
• SSL can effectively replace SL algorithms for tumor type diagnosis, requiring only a small labeled sample and less computational power.