Machine learning models for breast cancer prediction in the early stages

Abstract

Nowadays, cancer is considered one of the more harmful diseases in the newlineworld. It is one of the most common cancers that affect women and people newlineAssigned Female at Birth (AFAB). It leads to death, if not identified early. newlineHence early detection is essential to avoid death. However, manually detecting newlinebreast cancer is complex and requires more time. Therefore, an advanced newlinesupervised machine learning-based breast cancer detection and classification newlinemodel using the heuristic approach is developed in the first contribution. The newlinedeveloped model helps doctors to effectively diagnose breast cancer. Initially, newlinethe required data are collected from the Electronic Medical Records (EMR). newlineThen the collected data is pre-processed to remove the noises and enhance the newlinequality. After, pre-processing the data, they are inputted into the Mayfly newlineAlgorithm (MFO) to perform optimal feature selection. Then, the optimally newlineselected features are then given as input to the ensemble machine-learning newlinemodel for performing the breast tumor detection and classification task. The newlineGradient Boosting Model (GBM) is combined with the Decision Tree (DT) to newlinedevelop the ensemble model. Finally, the effectiveness of the developed newlineGradient Boosting Decision Tree-based Mayfly Optimization (GBDTMO) is newlinecompared with existing models to assess its performance in breast cancer newlineclassification tasks. Experimental results demonstrate the increased accuracy of newlinethe implemented GBDTMO model in identifying and classifying breast cancer. newlineHowever, this model is slower and it is complicated as it uses an ensemble newlinemodel for performing the detection and classification task. Therefore, a deep newlinelearning-based breast cancer prediction model is developed in the second newlinecontribution. The needed medical data are collected from the EMR. newline