Abstract:
Millions of homes are being equipped with smart devices (e.g. smart meters, sensors etc.) which generate massive volumes of fine-grained and indexical data that can be analyzed to support smart city services. In this paper, we propose to refine a model and introduce distributed learning of big data mining from multiple smart houses in a near real-time manner. This will help health applications to promptly take actions such as sending alert to patients or care providers. Furthermore, our work addresses building a health ontology model that automatically map discovered appliances to potential activities. This means we can competently train the system and increase the accuracy of sensing human activities. After the pre-processing stage in the data mining process, in the data classification phase, m5rules, Decision table, Random forest, random tree, Multilayer Perceptron (MLP), Farthest first and Logistic Regression (LR) algorithms have been used. The success evaluation of data mining classification algorithms has been admitted through the data mining program WEKA. Multilayer Perceptron algorithm has been the best algorithm with the highest success percentage in the program; Farthest first has been the algorithm which has the lowest success percentage in the program. This study has indicated that data mining can be a useful tool in the medical field, Doctors can be provided with daily activities of the patients in the progress of their treatment. The proposed mechanism, of this research uses the UMass Trace Repository which provides network, storage, and other traces to the research community for analysis. The result of accuracy values of WEKA classifying Algorithms are presented in this paper.
Keywords: Prediction, Big data, smart homes, smart devices, health care applications, Protégé, Medical Decision Support System
