Identification of the Risk Factors of Cervical Cancer Applying Feature Selection Approaches

A risk factor is any attribute that enhances the probability of getting an injury or disease. Cervical cancer has been considered as a frequent type of cancer among women. The attributes that are involved in enhancing the likelihood of cervical cancer are considered as the risk factors of this cancer. If it is possible to point out the risk factors of cervical cancer, the reason behind this disease can be understood easily. Thus, through proper treatment, mortality rate can be minimized and the quality of care can be increased. The aim of this study is to point out the risk factors of cervical cancer and find a model that provides good performance in predicting cervical cancer. In this research, Recursive Feature Elimination (RFE) and an ensemble approach based on Random Forests classifier have been used to identify the factors having much significance in predicting cervical cancer. Support Vector Machine (SVM), Multilayer Perceptron, and Logistic Regression have been used to evaluate the performance. SVM with an ensemble approach based on Random Forests classifier has outperformed than others with the accuracy of 91.04%, specificity of 91.94% , and AUC score of 0.89 for an independent test set.