Ensemble Method for User Activity classification in Ambient Assisted Living

Artificial Intelligence(AI) has become a global plat-form that allows objects in IoT to Interact and perform computations. The wide range of application areas of IoT are Smart Cities, Smart grids, Smart Supply chain and Ambient Assisted Living(AAL). These applications have challenges like tolerance to uncertainty,adaptiveness to the changing environment and improved trust among users. Thus, machine learning algorithms improve the performance of smart objects in various environment. The AAL environment deploys heterogeneous devices and sensors to capture various activities carried out through the daily by the individuals who resides in the smart home. In this work, an ensemble method using k-Nearest Neighbor(KNN), Decision Tree(DT) and Logistic Regression(LR)is proposed by investigating the performance of existing conventional supervised machine learning algorithms and selecting best model by considering the sensors features and improves the performance metrics. The work is evaluated using the benchmark ARAS (Activity Recognition with Ambient Sensing) dataset. The results are analysed using different parameters. The comparative analysis show that the proposed ensemble method gives accuracy of 76.28%.