Using Machine Learning and Feedback Filters for Localization in Ambient Assisted Living (AAL) Applications

Abstract

Machine Learning (ML) based Indoor Positioning Systems (IPS) are more efficient than other classical localization algorithms developed. Rather, efficiency these ML based IPS are easy to deploy in real environments. ML-based IPS initiates cognitive Location Based Services (LBS) in IoT. Among these LBSs, Ambient Assisted Living applications are curtailed. In this paper, we experiment with how to use ML classifiers in such an AAL application. During the experiments supervised classifiers Linear Discriminant Analysis Model, Quadratic Discriminant Analysis Model, Naïve Bayes Classifier Model, Decision Tree Classifier Model, and K-Nearest Neighbor Model were trained using an available Received Strength Indicator (RSSI) dataset to predict the location of a human. Algorithm Quadratic Discriminant Analysis provides a 25.88% misclassification error and 25.86% generalization error.