Identifying the most effective knowledge factors on HPV’s stigma among healthcare workers using feature selection models

Background and Aims: It is well known that stigma is a problem that affects numerous individuals,including HPV patients. The main objective of this paper is to determine the stigma associatedwith HPV by identifying the most correlated knowledge factors regarding HPV’s stigma inhealthcare workers using machine learning models.Method: A previously piloted and evaluated questionnaire of ۶۱ questions was filled out by ۵۵۲healthcare workers. The questionnaire included demographic, knowledge, and stigma question.Stigma questions’ answers were scored based on a Likert scale, and the total stigma score wascalculated for each case. Our data was split into train and test groups with proportions of ۲۰% and۸۰%. For both approaches, the total stigma score was used as the target variable and knowledgequestions and demographic data as the predictors. RapidMiner’s auto feature selection was utilizedin the first approach for Generalized Liner, Deep Learning, Decision Tree, Random Forest,Gradient Boost, and SVM models. All possible feature combinations were tested, and featureswere weighted. For the second approach, models including caret, Recursive Feature Elimination,Genetic Algorithm, Simulated Annealing, and XGBoost were applied to the data using R. Themost similarly voted and the most weighted features were later reported and compared betweenthe two approaches.Results: The train set included ۲۶۵ cases, and the test set included ۶۱. Our first approach evaluated۷۶,۲۹۰ models of six types of ML models and evaluated ۱۹,۳۹۲ feature sets; our top ۵ mostweighted features based on the correlation of six models were Faculty, Age, Occupation, ۱۷thQuestion (Using condoms can completely protect the individual from HPV), ۱۴th Question (HPVcan cause sterilization). Based on the second approach, the most similarly important features wereAge, Marriage Status, Occupation, Faculty, ۱۷th Question, ۱۴th Question, and Education Level.Conclusion: We have concluded that both of our approaches were able to detect questions andfactors that are most important regarding HPV stigma; our results can be used to create a muchsmaller questionnaire and help with channeling resources through the most effective areas to increaseHPV knowledge and decrees HPV stigma.

کلیدواژه ها:

Machine Learning ، HPV ، Stigma ، Vaccine