p03 / Preprint
Predicting Early Dropouts of an Active and Healthy Ageing App
arXiv:2308.00539 · 2023
In plain language
Machine learning predicts which users may stop using an active and healthy ageing application early.
Research context
Early adherence predictions can help experts identify where timely support may be useful.
Approach
Static and dynamic features are processed into seven datasets. Classifiers use SMOTE and ADASYN to address class imbalance.
Findings
Dynamic features improve classification, and oversampling produces a reported 10% improvement. The submitted methods earned first place in the IFMBE Scientific Challenge 2022.
Summary based on abstract.