Predicting Aggressive Behavior in Dementia Patients Using Text Classification with Word2Vec-LSTM

Abstract

Aggressive behaviour in dementia patients poses significant challenges for caregivers and healthcare providers. This study aims to develop and evaluate Long Short-Term Memory (LSTM) and Bidirectional LSTM (BiLSTM) models, integrated with word2vec embedding, for accurately predicting aggressive behaviour in dementia patients. Leveraging existing datasets containing pertinent information such as agitation levels and location, our models are trained to discern patterns indicative of aggressive episodes. Healthcare, a complex domain notorious for its diagnostic intricacies, stands to benefit greatly from such predictive analytics. We assess the efficacy of our models by comparing their predictive accuracy against established methodologies in dementia care. Furthermore, we investigate techniques to enhance model performance and discuss potential applications within clinical settings. This research underscores the utility of machine learning and deep learning in addressing critical challenges within healthcare, particularly in the realm of behavioural prediction in dementia care.