In Mayuge District, Eastern Uganda, adolescent girls and young women face significant challenges, including high rates of unwanted teenage pregnancies and HIV infections. To combat these challenges, our project aims to educate girls about the relative risk of sexually partnering with older men, empowering them to make informed decisions and reduce their vulnerability.
As part of this initiative, we have developed an innovative AI tool named MUWALA (meaning “girl” in Luganda) to identify high-risk factors among these girls. This tool focuses on emotional well-being, stress, and support to predict the risks of involvement in cross-generational sexual relationships, the risk of teenage pregnancy, and school dropouts. Users interact with MUWALA by filling out the online questionnaire, allowing the AI to analyze their responses and provide personalized risk scores. We trained a total of 150 girls from St. Joseph’s Primary School in sexual education classes. To inform the development of the AI tool, we collected data from 208 participants, including 117 in-school and 91 out-of-school girls and women, with signed consent forms.
The development of MUWALA involved a comprehensive workflow that started with cleaning the data using a CatBoost classifier, addressing missing responses, determining the importance of different factors, and then retraining the model to ensure optimal performance, utilizing metrics such as the F-1 score, SMOTE-ENC, and cross-validation for consistency across various test samples.
MUWALA was deployed using Django for the backend, a REST API for handling requests, and Next.js for the frontend. The tool has achieved an impressive accuracy score of 94.0% for pregnancy risk, 96.1% for dropout risk, and 70% for the risk of cross-generational sexual relationships. Our goal is to enhance MUWALA by developing it into a mobile application that enables offline access, widens its reach, and reduces response time.
By empowering young girls with timely information and support, we aim to create sustainable change in their communities.
