Cervical cancer risk assessment using machine learning integrated fuzzy MCDM methodology

Cervical cancer is entirely preventable if diagnosed at an early stage; however, the current rate of cervical cancer screening participation is not very adequate, and early detection approaches are still open and demanding. Evaluating the risk levels of potential patients in a practical and economic way is crucial to direct risky candidates to screening and establishing potential treatments to conquer the disease. In this study, a machine learning-integrated fuzzy multi-criteria decision-making (MCDM) methodology is proposed to assess the cervical cancer risk levels of patients. In this context, based on behavioral criteria obtained from the publicly accessible cervical cancer behavior risk data set from the UCI repository, the risk levels of patients are evaluated. The proposed methodology is established in three stages: In the first stage, using a machine learning technique, i.e., feature selection, the most effective criteria for predicting cervical cancer risk are selected. In the second stage, the criteria for importance through intercriteria correlation (CRITIC) method is used to assign objective importance levels to the criteria. In the third stage, the cervical cancer risk levels of candidate patients are prioritized using the technique for order preference by similarity to the ideal solution (TOPSIS) and, alternatively, the evaluation based on distance from the average solution (EDAS) techniques. The proposed methodology is developed in an interval-valued Pythagorean fuzzy atmosphere for quantifying the uncertainty in the nature of the problem. This study demonstrates that the feature selection algorithm can be efficiently utilized to determine the fundamental criteria of an MCDM problem and to aid in the early identification of cervical cancer.