Cervical cancer is one of the major important and growing causes of death in favor of middle-aged women in the growing countries, yet it is approximately totally unnecessary if precancerous lesions are detected and treated quickly. Recently there are few methods are introduced to classify cervical cancer accurately at present. In the classification of cervical cancer, choosing of important features is very challenging issue. To solve this issue, several feature selection algorithms has been introduced lately. Swarm based feature selection methods have been focused noticeably. Fuzzy Swallow Swarm Based Feature Selection (FSSBFS) have been introduced for optimal selection of cervical cancer features. Adaptive inertia weight and via integrating two fuzzy logic systems such as triangular and trapezoidal membership functions toward precisely compute the acceleration coefficients of FSS. These membership functions are introduced to FSSBFS for fuzzification; accordingly, its classification accuracy is improved and reduced computation time. In the FSSBFS algorithm risk factor features are chosen and then feed as input for classification. Improved Support Vector Machine - FSSBFS (ISVM-FSSBFS) is proposed to classify the malignant cancer samples. The cervical cancer samples are characterized by 32 risk factors and 4 target classes: Hinselmann, Schiller, Cytology, and Biopsy. From the results it demonstrated that the proposed ISVM-FSSBFS classifier is improved when compared to SVM and Multilayer Perception Classifier (MLP) classifiers.
Volume 11 | Issue 9
Pages: 77-88
DOI: 10.5373/JARDCS/V11I9/20192917