Deep Recurrent Q Reinforcement Learning model to Predict the Alzheimer Disease using Smart Home Sensor Data

Abstract

Abstract Early detection of alzheimer disease is an essential process of the elderly persons when they are with the mild cognitive impairments. This work adapts the smart home based alzheimer disease detection by recording the daily activities of the resident’s equipped with the sensor devices in their appliances as smart home. Using sensor data based clinical assessment of alzheimer patient is very complex as it is uncertain to predict such vague monitoring and understanding its deep characteristics. To acquire the deep characteristics of sensor dataset unlike the existing conventional supervised learning paradigm, this paper constructs an optimized self-learning model known as reinforcement learning process. This proposed model Deep Recurrent Q Learning based Reinforcement Model (DRQLRM) Process comprised of two stages they are pretraining using recurrent neural network to overcome the overfitting and fine tuning of the parameters involved in deep neural network and Deep Q learning process is used as reinforcement approach which uses the agents rewards and penalties to determine the optimal agents and their experience to predict the unknown pattern in the uncertain condition in effective way. The proposed model produced better result in CASA smart home test bed for investigating the presence of alzheimer