Classification Method of Indoor Environmental Data based on Internet of Things

Aiming at the multi-source heterogeneous characteristics of data collected by Internet of Things wireless sensors, we investigate an IoT data classification method based on Improved Multi-Kernel Learning Support Vector Machine (IMKL-SVM). Kernel function types and parameters are selected mainly according to rule of thumb in traditional multi-kernel learning method. Our improved method has two steps in determining types and parameters of the kernel function. First, the primarily types and parameters of the kernel function is determined by a cross-validation method. And then the types and parameters of the multi-kernel function is optimized by SVM. The experiment designed two sets of data including four sensor data: temperature, humidity, light and atmospheric pressure. One group data was labeled as four types: morning, mid-afternoon, evening and night. And another group data was labeled as three types: day, evening and night. The classification accuracy are presented using the improved method, single-kernel SVM and traditional MKL-SVM for the two group data. Experiment results indicate that the IMKL-SVM method achieves higher classification accuracy for IoT data with multi-source heterogeneous characteristics.