Sentiment Analysis on the Level of Adoption of Health Applications for Pregnant Women: Study Case with LDA and SVM

Technology adoption in Indonesia’s health sector has significantly improved access to healthcare services, particularly for maternal health. However, the uptake of health applications among pregnant women remains uneven and faces several challenges. Key barriers include concerns over data security and privacy, limited accessibility due to varying levels of digital literacy and infrastructure, and sub-optimal feature design that does not fully meet user needs. This study analyzes a large dataset of 74,599 Google Play reviews of maternal-health apps, focusing on three critical adoption factors: security, social influence, and health-related content. This study employs a hybrid machine learning approach combining a Support Vector Machine (SVM) classifier with a Latent Dirichlet Allocation (LDA) topic model, achieving an accuracy rate of 87%, which surpasses the performance of earlier convolutional neural network (CNN)-based methods combined with TextBlob sentiment analysis. The topic modeling reveals important user concerns and preferences, such as the need for stronger in-app privacy controls, richer localized resources tailored to Indonesian contexts, and smoother onboarding processes to enhance overall user experience. To our knowledge, this research is among the first in Indonesia to integrate LDA and SVM techniques for maternal-health sentiment analysis, offering a scalable and effective framework that can inform developers and policymakers in designing more user-centered digital health solutions.