Cervical Cancer Screening using Data Mining Technique

Abstract

Cervical cancer is one of the most common cancers in females these days. Previous screening diagnosis of cervical cancer has been done by several methods. One method is to check the medical history, HPV high-risk type testing, body fluids, PAP smear, and tissue biopsy. In this paper, we proposed a cervical cancer screening diagnostic method by using data mining with Ant-Miner Algorithms. The objective was to search the data mining techniques to create a cervical cancer screening model of efficiency in the classification and feature selection for the data mining method through a correlation-based approach.    These experiments on medical datasets (There are 32 attributes, 4 classes with 858 samples) showed that Correlation-based Feature Selection (CFS- good feature sets contain attributes that are highly correlated with the class) rapidly identifies and screens unrelated, humdrum, and missing features, and identifies relevant features as long as their relevance does not strongly depend on other features. CFS helps by providing a smaller number of features with the high performance of cervical cancer screened by accuracy and precision. The results show that age, number of sexual partners, first sexual intercourse, number of pregnancies, hormonal contraceptives, and IUDS are the main predictive features for cervical cancer.        The screening model of total classes showed a high average accuracy of 94.68% with an average precision of 93.78%. When considered by the type of class the results are as follows: the accuracy of the Hinselmann class was 93.26%, with a precision of 90.00%, the accuracy of the Schiller class was 90.86%, with a precision of 95.24%. The accuracy of the Cytology class was 96.26%, with a precision of 92.10% and the accuracy of the Biopsy class was 98.35%, with a precision of 97.78% respectively. Data mining with the Ant-Miner Algorithm has shown to be advantageous in handling a cervical cancer screening diagnostic assignment with excellent performance.

Keywords: Cervical Cancer, Correlation of Data, Data Mining