Machine Learning and Feature Selection for Breast Cancer Prediction

One of the most prevalent and fatal tumors that impact women globally is breast cancer. Traditional diagnostic methods, while effective, can be costly. The goal of this research is to improve the precision and effectiveness of breast cancer detection by combining feature selection techniques with machine learning models. Seven machine learning models were trained and assessed using the Wisconsin Diagnostic Breast Cancer (WDBC) dataset in conjunction with three feature selection strategies: filter method, univariate selection (SelectKBest), and embedded method (Random Forest importance). Experimental results show that neural networks achieved the highest performance when using all features, while ensemble models performed best when used with filter feature selection. The study found that the choice of feature selection method should be aligned with the nature of the model, and combining suitable selection strategies with machine learning models can significantly enhance diagnostic performance. This approach can reduce misdiagnosis and improve early treatment outcomes.