Spatio-Temporal Abnormal Behavior Prediction in Elderly Persons Using Deep Learning Models.

Area of Science:

• Artificial Intelligence
• Gerontology
• Health Informatics

Background:

• Monitoring elderly individuals' behavior is crucial for health and safety in smart home environments.
• Abnormal behaviors during daily activities can signal potential health issues requiring intervention.
• Maintaining independence for the elderly necessitates reliable health monitoring systems.

Purpose of the Study:

• To investigate and compare the performance of various deep learning models for identifying and predicting abnormal behaviors in elderly persons.
• To evaluate the efficacy of Long Short Term Memory (LSTM), Convolutional Neural Network (CNN), CNN-LSTM, and Autoencoder-CNN-LSTM models.

Main Methods:

• Utilized temporal and spatial data collected over time to train deep learning models.
• Experimentally evaluated the performance of different model architectures and hyperparameter tunings.
• Tested models on two public datasets for identifying and predicting abnormal behaviors in smart homes.

Main Results:

• The study presents an experimental evaluation of the selected deep learning models' performance.
• Accuracy was the primary measure for evaluating the models' ability to detect abnormal behaviors.
• Different model architectures and hyperparameter settings were considered to optimize performance.

Conclusions:

• Deep learning models show promise in accurately identifying and predicting abnormal behaviors in elderly individuals.
• The findings support the development of advanced health monitoring systems for the elderly in smart environments.
• Further research can refine these models for enhanced accuracy and broader application in elder care.