Recall-based Machine Learning approach for early detection of Cervical Cancer

With frequent advancements in development of algorithms and need to incorporate them with clinically synthesized medical information is the paramount of modern-day bioinformatics. This aspect of computational study is of great healthcare significance as deduced results could be farfetched to generalize conclusions in regular medicine practice and diagnosis thus fastening up the process of detection. This paper tries to generalize cervical cancer detection approach with random forest regression technique. Unlike other papers which focus on accuracy and precision, this paper emphasizes on recall-based approach and beneficial tenets this approach over former ones. Four diagnostic tests used for early stage detection of Cervical cancer are Hinselmann's test, Schiller's test, Biopsy and Cytology. Each test is studied individually and analysis was made on the basis of confusion matrix, recall score and receiver operator curve (ROC). The basic aim during the entire development was to achieve higher recall scores with reduced false positive values.