Outlier detection in sensed data using statistical learning models for IoT

Internet of Things (IoT) devices are composed of millions of sensors that continuously sense environmental parameters which are effectively fused to gain insights on a particular area or region based on which desired actions are taken. However, before any fusion of the data takes place, checking the quality of data is of paramount importance as these data may also be contaminated with outliers which limit their overall efficiency. Outliers or in other words, any abrupt change in values can be either due to a drastic change in environment (Event) or because of any malfunctioning of the sensor (Error). This paper introduces an IoT architecture to detect the occurrence of both Error and Event in a forest environment with the help of four statistical models, i.e., Classification and Regression Trees (CART), Random Forest (RF), Gradient Boosting Machine (GBM) and Linear Discriminant Analysis (LDA). We take into account the spatial and temporal dependencies of the data as well. Simulation results show that the models can effectively detect both types of outliers with upto 100% accuracy for Error detection and upto 98.51% in case of Event detection. Also, the results provide evidence that using spatial and temporal parameters in event detection can have a positive influence on the prediction accuracy which was observed to increase from 81.68% to 96.53% using RF. Among the four models, RF is seen to outperform the others.