Big Data Analytics in IoT Ecosystems

Internet of things (IoT) ecosytems stew (1) connect few, if any, of many different devices in many different places, creating many degrees of heterogeneous data. Big Data that resulted from IoT needed to be used as information, and was popularized as the most important tools of big data analytics in IoT. This research is to solve the IoT data high dimensionality and complexity problems by the combination of the support vector machine (SVM) and the recursive feature elimination (RFE), and the two step solution is it. Features of the input and characteristic extracted from the IoT system are used to classify and/or predict IoT system behaviors by SVM, and RFE is used to process or filter the huge datasets to eliminate the unimportant feature, with this eliminating the noise that will not hamper accurate performance of the models. We find that the hybrid approach improves speed up and scalability and these models make the model realistic to use for, e.g., predictive maintenance, anomaly detection, real time decision making. Based on this, from the presented results we can infer that introducing more involved machine learning methods could help us getting better optimization of the utilization of IoT based Big Data.