Abstract

Cervical cancer is one of the most widely recognized gynecologic malignancies on the world and it is demanding since this malignant growth happens with no signs. As per World Health Organization (WHO), cervical cancer is the fourth most recurrent disease which is higher death rate that influenced women everywhere in the world. It has demonstrated that early discovery of any cancer when followed up with suitable diagnosis and treatment can expand the patient survival rate. But the existing techniques have problem with imbalanced dataset and feature selection-based classification accuracy. To conquer the previously mentioned issues, the existing strategies are analyzed different procedures of data mining and feature selection techniques which can be applied to bring out hidden information from the cervical cancer dataset. In this review, classification process and feature selection-based classification are performed to improve the given cervical cancer dataset accuracy significantly. In the classification process, the imbalanced data and redundant features are not handled effectively. Hence the feature selection-based classification is required to improve the cervical cancer classification accuracy. This survey is also analyzed the merits and shortcomings of each method applied to application. The comparative analysis is done using various classification techniques like Support Vector Machine (SVM), K Nearest Neighbor (KNN), Convolution Neural Network (CNN) and Synthetic Minority Oversampling Technique + Random Forest with Recursive Feature Elimination (SMOTE+RFE+RF) approach. The experimental result shows that the SMOTE+RFE+RF approach provides better performance in terms of higher accuracy, specificity, Positive Predicted Accuracy (PPA) and Negative Predicted Accuracy (NPA) and sensitivity rather than the other existing methods.