Multiclass Predictor and Diagnoses of Breast Cancer Using Machine Learning

Metastatic Breast Cancer is the most prevalent and life-threatening diseases worldwide, affecting millions of women and men each year. Among all the most important factors that significantly influence the diagnosis and treatment an early stages are vital for reducing the fatality rate in the fight against it. Breast cancer subtype classification is a multi-class classification problem that can be handled using computational methods. There are numbers of techniques as thermography, ultra-sound, mammography and magnetic resonance imaging (MRI) for the detection of cancer tumor. By combining these with machine learning techniques and automated diagnostic techniques, the potential for achieving highly accurate diagnoses with minimal false outcome. In this paper a novel intelligent integrated diagnosis system is proposed using classification models including Decision Tree Classifier, Naive Bayes Multinomial classifier, Support Vector Machines, K-Nearest Neighbor, XGBoost and Random forest using features selected at different threshold levels. The integration of machine learning techniques with clinical and histopathological data offers a powerful approach to analyzing complex patterns and relationships, thereby enhancing the ability to identify patients at high risk of metastasis in a timely manner. The performance and results evaluation shown an accuracy that varies for the models amongst five proposed approaches gives 94% to 99%, with a precision value of 95% to 98%, sensitivity from 96% to 99 %, and with specificity from 93% to 97%. This work will always support to the patient and doctors to take help of this model to predict the complexity of the breast cancer and its type to reduce the risk of this life threatening diseases.