A comparative analysis of hyper-parameter tuned supervised machine learning algorithms on breast cancer prediction
Taki Hasan Rafi · 2020
In recent years, breast cancer is one of the expansive diseases among various sorts of tumors in women. Around 20 million women are determined to have breast cancer globally. It is normal that out of 8 ladies, one lady would get affected by breast tumor. In this specific situation, early diagnosis of the disease can increase the survival rate of the patients. There are some different ways to diagnose. In this paper, author attempted to tackle this issue specifically by machine learning applications. It has been executed using several supervised learning algorithms, for example, Support Vector Machine (SVM), Logistic Regression (LR), Random Forest (RF) and K-Nearest Neighbor (K-NN). These algorithms show a noteworthy accuracy in performance over the breast cancer dataset. Logistic Regression has an accuracy of 94.50%, SVM has that of 97.58%, K-NN has that of 96.70% and Random Forest has that of 92.39%. Hyper-parameter tuning is a way of finding best optimal parameter to learn the machine learning algorithms. However, after hyper-parameter tuning of each model, SVM has revealed the most noteworthy accuracy among other models. Author has decisively compared the proposed model with different past research models.