Chong, PeterGutierrez, JairoMohan, Deepika2025-08-272025-08-272025http://hdl.handle.net/10292/19730Senior citizens are one of the significant proportions of healthcare service users, with approximately 12% of public sector, primary care, and hospital services utilized by this demographic. Senior Citizens living alone, without adequate monitoring, are at a higher risk of severe consequences from sudden falls caused by slips, trips, or underlying health conditions. In New Zealand, approximately one-third of people over 65 years fall each year. Among those who experience a fall, 22–60% sustain injuries, 10–15% suffer serious injuries, 2–6% experience fractures, and 0.2–5% sustain hip fractures. People in their 80s and above are particularly at high risk of falling. Overall, half of all ACC claims and costs in this age group result from falls, amounting to estimated costs of NZ$443 million annually. The majority of these incidents go unreported in time for emergency interventions, causing critical delays in treatment. This affirms the necessity of trustworthy and affordable e-health technologies to support independent living among older adults. In the modern era, Artificial Intelligence (AI) and Machine Learning (ML) have significantly impacted daily life, proving instrumental in various domains. In this context, AI can serve as a vital companion for older adults, providing continuous health and behaviour monitoring and early fall prediction capabilities. The focus of this thesis is on leveraging AI to forecast potential falls in older adults by identifying abnormalities in their health and behavioural patterns. The proposed approach termed the Co-operative AI model for future fall prediction in the Elderly, employs Fuzzy logic and deep learning networks and algorithms such as Deep Belief Networks (DBN) for the development of two AI models that are combined using a Meta-Model to provide the Future Fall Risk Prediction Outcome. AI-1 model utilizes vital sign parameters such as Blood Pressure, Heart Rate, and Oxygen Saturation as input and employs Fuzzy Logic to predict potential fall risk in older adults. The AI-2 model, on the other hand, uses Activities of Daily Living (ADLs) such as Sitting, Standing, Walking, Running, and Jumping as input parameters and employs a Deep Belief Network (DBN) for fall risk prediction. The input dataset for both AI-1 and AI-2 models was collected from PhysioNet – a Public Repository. The outputs from AI-1 and AI-2 models are then integrated using a Meta-Model which uses a Random Forest with continuous learning features, which provides a comprehensive prediction of impending falls in older adults. The thesis has the following main contributions: • A detailed background analysis was done in terms of health, behaviour, and environmental factors, and the critical parameters that lead to an abnormality in older adults were analyzed. • The system architecture consists of 3 AI models (AI-1 model, AI-2 model, and Meta-model) where each model works independently to assess fall risks with different factors, then works collaboratively by learning from each other to generate a promising final fall risk prediction result. • Data is collected from the public repository for both the AI models. The system continuously monitors health and behavioural parameters, processes the information through deep learning models, and predicts early indicators of fall risk. By analyzing these parameters in real time, the model effectively identifies abnormalities that signal potential fall risks. In a nutshell, this thesis proposed the research findings and model development which highlights the significant potential of advanced deep learning techniques to evaluate the severity of risks, to improve future fall risk prediction, and finally to predict and reduce fall occurrences. This innovative technology provides a transformative approach, enabling timely interventions that improve the safety and well-being of elderly people while potentially saving lives.enThesisOpenAccess