Performance Analysis of Selected Machine Learning Algorithms for the Detection of Cervical Cancer Based on Behavioral Risk Dataset

ABSTRACT:

Past studies have reported that cervical cancer is the second most common cancer among women in the world. Aside medical diagnosis and treatment, machine learning algorithms have been used widely for the classification of cervical cancer. These machine learning techniques have been found promising for cervical cancer identification by making use of relevant cervical cancer datasets. Unlike many similar studies that focus on using datasets with gene expression or histological images for classifying presence of cervical disease in patients, this study used behavioural dataset for building two predictive models. This work also focuses on investigating how scaling of the features as well as selection of promising ones can impact the predictive ability of the classification models. Our investigated the performances of the supervised algorithms in the chosen cervical cancer behavioural risk dataset. The behavioural risk dataset is chosen for the study in order to advance cervical cancer studies and establish how the features in such dataset can be used to train and test learning classifiers for cervical cancer identification. In the article, Gaussian Naïve Bayes (GNB) and Logistic regression algorithms were used for building the predictive models. The study obtained the performances of GNB-based model as: accuracy -0.98, precision-0.97, recall-0.96 and fmeasure 0.94 respectively. Similarly, Logistic Regression-based model achieved accuracy of 0.90, precision of 0.87, recall of 0.86 and f-measure is 0.83. Thus, experimental results showed that GNBbased model outperformed its logistic regression
counterpart

Keywords: Behavioral Health, Cervical Cancer, Machine Learning Prediction, Behavioral. Features.