Stacked model verification algorithm in fall detection

This study aims to address the issue of fall detection in the elderly, using a wearable device-based fall detection system. Acceleration data from the wearable devices was collected to construct the dataset. After applying certain algorithmic processing to the dataset, it was found that the generalization ability of a single machine learning algorithm was generally poor. In order to improve classification accuracy, we attempted to use ensemble learning algorithms to train and validate the fall detection dataset. By employing a stacked model and utilizing different combinations of base learners and meta-learners, our model achieved an average accuracy of 99.42% on the validation set after 0.2 cross-validation, greatly enhancing the model’s high performance and strong generalization ability.