Prognostic Evaluation of Cervical Cancer Using Various Classifiers

Abstract

Cervical HPV (human papillomavirus) infection is almost invariably the prominent reason for cervical cancer. The most prevalent malignancy in female is cervical cancer and it substantially affects the risk factor for mortality. Cervical cancer is bound by a wide range of risk factors. Analyzing risk factors from each patient's medical history and screening outcomes, we created a prediction model in this study to forecast the course of cervical cancer patients. Early detection and diagnosis are the most effective methods of lowering the incidence of cervical cancer globally since it is a condition that is largely preventable. Early detection of the signs of this gynecologic disease can be challenging, particularly in nations without screening systems. Machine learning techniques can be utilized to identify malignant cancerous cells when cervical cancer was first discovered. We created a machine-learning technique that can handle massive volumes of data concurrently with greater accuracy. Based upon many parameters, such as age and lifestyle, can forecast a person's likelihood of developing cervical cancer. A Kaggle data repository Cervical Cancer dataset was retrieved, it had eight hundred and fifty-eight unique data sets from distinct cases involving thirty-six variants comprised of diverse risk and protective factors. And it had four types of attributes: Cytology, Hinselmann, Biopsy, and Schiller. Using these class attributes as a basis, this dataset was divided into four groups. To detect cancer and facilitate prompt identification, ML classification algorithms like Decision Trees (DT), Logistic Regression (LR), Random Forest (RT), Support vector machines (SVM), AdaBoost, and artificial neural networks have been implemented. After gathering the data, in order to investigate the prevalence of characteristics and interconnections between numerous unrelated parameters and the propensity to acquire cervical malignancy, we first carried out a descriptive statistical study. According to the study's inferences, appropriate network architecture, categorization, and ML algorithms are capable of properly and effectively detecting Cervical Cancer in its early stages utilizing diagnostic information. Eventually, while comparing the acquired outcomes, we scrutinized that SVM, AdaBoost, DT, and LR algorithms had exhibited hundred percentage of performance, while RF and ANN algorithms seemed to have ninety-nine percentage of performance. Since the goal of our research was an early cancer forecast. Than such, we suggest that future research should take into account expanding patient histories, for example, by integrating primary health information before hospital referral.