An Enhancement in Accuracy for Breast Cancer Prediction Using Machine Learning and Deep Learning Model

As we are all aware, the population is growing rapidly in this modern period, placing a demand on the healthcare system to identify illnesses in the populace. This creates a major issue for eradicating large-scale chronic diseases. Breast cancer is one of them. After lung cancer, breast cancer is listed as one of the most widespread cancers. A efficient and fast response are provided by an automated illness detection system using different machine learning models, which also lowers the chance of fatalities while helping medical professionals in disease identification. Here, we have compared six machine learning approaches named K-nearest neighbors (KNN), support vector machine (SVM), artificial neural networks (ANNs), random forests (RFs), parallel adaptive local hyperplane (PALH) and logistic regression (LR). The overall performance is evaluated with reference to precision, accuracy, specificity, sensitivity, negative predictive value, false-positivity rate, false-negativity rate, Matthews Correlation Coefficient and F1 score. The results reveal that the PALH obtained the highest score in accuracy, specificity, precision and F1 score of 98.68, 96.34, 99.10 and 0.9919, respectively, whereas 98.57, 96, 97.82% and 0.9890 accuracy, specificity, precision and F1 score are obtained by ANN, respectively. We have tested the algorithms on Breast Cancer Wisconsin (Original) dataset. The PALH has advantages over deep learning models (ANN) like it requires less computational power, is easily optimizable, easy to interpret and deploy. Thus, PALH has overcome the limitations of ANN while achieving a better accuracy, precision and F1 score.