Machine Learning and Deep Learning Approaches in Breast Cancer Survival Prediction Using Clinical Data.

Area of Science:

• Oncology
• Biomedical Informatics
• Data Science

Background:

• Accurate breast cancer survival prediction is crucial for effective treatment planning.
• Traditional statistical and machine learning models have limitations in prediction accuracy.
• Deep learning offers potential for enhanced breast cancer prognosis models.

Purpose of the Study:

• To evaluate and compare machine learning and deep learning models for breast cancer survival prediction.
• To identify key features influencing breast cancer survivability.
• To assess the accuracy of different algorithms in predicting patient outcomes.

Main Methods:

• Utilized a dataset of 4,902 breast cancer patient records from the University of Malaya Medical Centre Breast Cancer Registry.
• Applied and compared various machine learning classifiers: Multilayer Perceptron (MLP), Random Forest (RF), Decision Tree (DT), and Support Vector Machine (SVM).
• Investigated the impact of different features, including tumor size, on prediction accuracy.

Main Results:

• Multilayer Perceptron (MLP) achieved the highest accuracy at 88.2%, followed by Random Forest (RF) at 83.3% and Decision Tree (DT) at 82.5%.
• Support Vector Machine (SVM) showed a prediction accuracy of 80.5%.
• Tumor size was identified as the most significant feature for predicting breast cancer survivability.

Conclusions:

• Both machine learning and deep learning approaches demonstrate significant potential for accurate breast cancer survival prediction.
• Model performance is influenced by factors such as parameter tuning and data preprocessing techniques.
• These predictive models can aid in optimizing treatment protocols and improving patient outcomes.