Predicting knowledge of cervical cancer screening among reproductive-age women in sub-saharan africa using machine learning algorithms.

Sub-Saharan Africa faces twice the incidence and up to fifteen times the fatality rate of cervical cancer compared to developed countries. Screening coverage remains low at 7-25.3%, far below the WHO target of 70%. Increasing women's knowledge about cervical cancer screening is crucial for improving uptake and reducing this high burden. The study was conducted among 50,584 reproductive-age women across seven Sub-Saharan African countries. A combination of filter and wrapper methods was applied for feature selection. Data preprocessing and management were performed using Stata version 17 and Python (Colab, version 3.10.2). MinMaxScaler, StandardScaler, and RobustScaler were applied to normalize variable ranges. One-hot encoding was used for nominal categories, and ordinal encoding was applied for features with an inherent order. The dataset was split using an 80; 20 (40,467: 10,117) ratio for training and testing. Eight algorithms were selected for model training and development, including Decision Tree, Random Forest, K-Nearest Neighbors (KNN), Artificial Neural Network (ANN), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting (LightGBM), Adaptive Boosting (AdaBoost), and Gradient Boosting (GB). Hyperparameter tuning was performed for XGBoost and LightGBM. Model performance was evaluated using Accuracy, AUC, F1 Score, Recall, and Precision. The XGBoost algorithm outperformed the other models, achieving an accuracy of 84%, an AUC of 82%, an F1 score of 83%, a recall of 82%, and a precision of 84%. Overall, 56% of reproductive-age women demonstrated good knowledge of cervical cancer screening. HIV status, educational status, media exposure, maternal age, health status, antenatal care attendance, wealth status, occupation, autonomy, place of delivery, distance to health facility, marital status, and fertility were identified as the top predictors of knowledge of cervical cancer screening. Knowledge of cervical cancer screening was low, at 56%, compared with the WHO target of 70%. Strengthen community-based educational programs through mass media, reinforce health education during antenatal care (ANC) follow-up and HIV programs, and implement strategies that promote women's autonomy, education, and economic empowerment are recommended.