Semi-supervised data stream analytics with balanced recognition performance and processing speed

For the upcoming IoT (Internet of things) era, plethora of data from a variety of sensors needs to be processed on a real time basis for improving system responsiveness. Due to the increasing modality of sensors, data streaming analytics to deal with high dimensional data becomes a critical ability. In addition, concept drift also needs to be addressed since an IoT-enabled environment is dynamic in nature. To address the above issues, this study proposed a semi-supervised algorithm for data streaming analytics under concept drift to achieve a tunable balance between the processing speed and the recognition accuracy based on the user needs or application characteristics. In the experiment using a KDD dataset, we found that the proposed algorithm does achieve an expected and promising result.