An Internal Clustering Validation Based Fitness Approach for Meta-Heuristic Diagnosis of Cervical Cancer
M. Kerem Un, Mustafa Güven, Çağlar Cengizler, Şeyda Erdoğan, Aysun Hatice Uguz · Frontiers in Signal Processing · 2020
This paper presents an utilization of data clustering with genetic algorithm (GA) approach.Proposed meta-heuristic clustering approach relies on genetic operators and accepts Calinski-Harabasz (CH) measure as fitness criteria where each individual represents a final judgement about existence of malignancy on set of cervical cells.It was aimed to evaluate the performance of fitness criteria on detection of malignancy where classification is performed on salient morphological features.Preferred fitness criteria measures the ability of individuals in a population to form appropriate clusters for normal and abnormal cell samples.Feature space includes data extracted from the previously segmented cervical cell images.Proposed approach is examined with two data sets which contains malignant and healthy cell samples.Preliminary results has shown that preferred fitness criteria for the classification is promising and the presented utilization of GA based clustering approach with CH criteria has a better clustering performance compared to conventional clustering methods.