Leveraging Naive Bayes for Enhanced Survival Analysis in Breast Cancer

Abstract

The study aims to predict breast cancer survival using Naïve Bayes techniques by comparing different machine learning models on a comprehensive dataset of patient records. The main classification groups were survival and non-survival. The objective was to assess the performance of the Naïve Bayes classifier in the field of data mining and to achieve significant results in survival classification, aligning with current academic research. The Naive Bayes classifier attained an average accuracy of 91.08%, indicating consistent performance, though with some variability across different folds. Conversely, Logistic Regression achieved a higher accuracy of 94.84%, demonstrating proficiency in recognizing instances of class 1, yet encountering challenges with class 0.The Decision Tree model, with an accuracy of 93.42%, exhibited similar performance patterns. With an accuracy of 95.68%, Random Forest surpassed the Decision Tree. Nonetheless, all models encountered challenges in accurately classifying instances of class 0. The Naive Bayes algorithm was juxtaposed with K-Nearest Neighbors (KNN) and Support Vector Machines (SVM). Future research aims to enhance prediction models with novel methods and tackle the challenge of accurately identifying instances of class 0.