Predictive Breast Cancer Learning Model for Selected Features: Comparative Analysis

Former predictions of any disease can be curable with doctors’ attention. Most of the time, early diagnosis fails due to a lack of people and doctor attention. According to WHO, early diagnosis of a disease can be cured before it becomes chronic. Breast cancer is a disease that can be cured if it gets identified at an early stage of cancer or before it starts spreading throughout the body parts. This paper implemented different feature selection methods with different machine learning classifiers. The breast cancer dataset from the UCI repository is partitioned into different train: test split ratios. Further learning models are measured based on statistical parameters like accuracy, sensitivity, specificity, precision, recall, and F1-score. Backward elimination and LASSO regression with SVM at 80:20 train: test split ratio produces the best results, and information gain with Naive Bayes at 70:30 split ratio gives the third best result.