Prediction of Cervical Cancer Using Boosting Techniques.

Ramoni Tirimisiyu Amosa¹*, Adekiigbe Adebanjo¹, Olawale Olaniran Kayode², Fabiyi Aderanti Alifat¹, Olorunlomerue Adam Biodun¹, Oluwatosin Adefunke Oluwatobi¹, Adejola Aanu Adeyinka & Fakiyesi Favour^1
1Department of Computer Science, School of Applied Sciences, Federal Polytechnic Ede, Osun State, Nigeria.
²Department of Science Laboratory Technology, School of Applied Sciences, Federal Polytechnic Ede, Osun State, Nigeria.*Corresponding Author
DOI: https://doi.org/10.51584/IJRIAS.2023.8605

Received: 06 May 2023; Revised: 02 June 2023; Accepted: 06 June 2023; Published: 03 July 2023

Abstract: Cancer of the cervix, commonly called cervical cancer, is a type of cancer that develops in the cells of the cervix, which is the lower portion of the uterus that attaches to the vagina. It hardly shown symptoms in its early stage. To detect the disease, regular is required, however larger population of women not aware of this approach while many shy away and refuse to take the test. Hence cervical cancer spread like wild fire among women and being the most common cause of cancer disease it result to untimely death among women in our society today. In this research, the performance of a few sophisticated ensemble models, such as Bagging Classifier and Adaptive Boosting (AdaBoost) Classifier, is shown for the purpose of predicting a diagnosis of cervical cancer based on recorded cancer risk factors and target variables. Accuracy, sensitivity, and specificity were the measures that were used in the evaluation of the models. Python library was adopted for the classification and the cervical cancer dataset used for the experiment was acquired from UCI (University of California at Irvine), the classification was carried using voting approach by combining three classifiers: Decision Tree (DT), K-N Neighbour(KNN) and Random Forest (RF). The results indicated that the proposed model was highly accurate in predicting the risk of cervical cancer, with 119 instances classified as ‘class zero’ and only three instances classified as ‘class one’ based on the predictions.

Keywords: Accuracy, Cervical Cancer, Experiment, Machine Learning (ML), Model, AdaBoost Classifier, Bagging Classifier.

I. Introduction

Cancer is a perilous illness characterized by the abnormal growth and division of cells, which occurs in a disorderly and uncontrolled manner, disregarding the normal regulations of cell division. (Hejmadi, 2014). Cancer is caused by the abnormal growth of cells in a specific part of the body, which multiply uncontrollably. It is a significant cause of death worldwide, with approximately 8.2 million people dying from cancer each year, accounting for 13% of all deaths globally. Screening services for cancer are not widely available, particularly in underdeveloped countries, with only 26% reporting screening services available for the public in 2017. While 90% of developed countries offer treatment services, less than 26% of low-income countries have access to such services. It is estimated that the number of cancer cases will increase to 22 million by 2030. (Ferlay et al., 2021). Lung and breast cancer are responsible for millions of premature deaths among women, but cervical cancer is considered to be the most perilous because it affects only females. The female reproductive system includes the cervix, uterus, vagina, and ovaries, and cervical cancer develops in the opening of the uterus from the vagina, which is known as the cervix. HPV, often known as the human papillomavirus, is the virus responsible for the development of cervical cancer (Shah & Itzkowitz, 2022)
Cervical cancer is more prevalent in low- and middle-income countries, as reported by Siegel (2018). Screening is crucial for detecting cervical cancer. An effective screening test should be minimally invasive, easy to perform, acceptable to the patient, affordable, and capable of diagnosing the disease in its early stages when treatment is most effective (Cohen et al, 2019). The four main screening methods for the disease are; cervical cytology, biopsy, Schiller test, and Hinselmann test (Desai et. al., 2021). Researchers have used decision tree for the prediction and some other approaches (Fahad, 2019; Keller et al., 2019; Hejmadi, 2014; Bray et al., 2018). However, previous results shows that further investigation on how to improve the result and accuracy of the prediction is imperative.

Full Text  PDF