An Ai Based Application for Cancer Diagnosis - An Emperical Analysis

Abstract

Cancer is one of the most dangerous diseases that may strike a person. Global mortality could occur as a result of the late diagnosis. Profuse research now calls for the early detection and start of treatment for the majority of cancer types as it helps with patient medical care. Lung cancer, liver cancer, breast cancer, and cervical cancer are four significant cancers that we have taken into account in this work. It does a comparison study between numerous current algorithms, including Linear Regression, Logistic Regression, Support Vector Machine (SVM), K-nearest neighbors (KNN), Naive Bayes, Gradient Boosting, Xtreme Gradient Boosting (XGBoost), and Random Forest based on all the datasets gathered for the prediction of the sorts of malignancies we've taken into consideration. The indicator that is used for comparing the algorithms is the prediction accuracy value. After performing the comparative analysis, we saw that different algorithms performed better on different datasets. On the breast cancer dataset, Logistic Regression performed best with an accuracy of 99.6%, on the Lung cancer dataset, K-nearest neighbor performed best with an accuracy of 97%, on the Cervical cancer dataset, Random Forest performed best with an accuracy of 99.2% and on Liver Cancer, Random Forest performed best with an accuracy of 73.73%. This will help in further research since to our best knowledge there haven't been many research papers that have compared the most widely used algorithms for cancer prediction models.