Superimposed Rule-Based Classification Algorithm in IoT

The application of the Internet of Things (IoT) in agriculture captures an enormous amount of data for decision making. However, hardest-to-detect abnormal data points that are transmitted can be harmful if not detected at an earlier stage. This paper presents an application of the Superimposed-Rule Based Classification Algorithm (SRBCA) using IoT in agriculture that lowers the false positive in anomaly detection by training one-class dataset of a ground-truth collection of agricultural sensor readings and evaluating using the combination of ground-truth and synthetic balanced test set. The CRoss-Industry Standard Process for Data Mining (CRISP-DM) methodology was used as a guide in the development of the study. The SRBCA was developed to detect conditional anomalous instances. The model was tested with one (1) year daily collection of environmental sensors. This algorithm considers the behavior and indicator features for anomaly detection. Moreover, a confusion matrix is presented showing the accuracy of the result of the SRBCA compared with One-Class Support Vector Machine (OCSVM) and its types which were considered as the closest prior art of the algorithm. The experimental results show that SRBCA performed better in identifying conditional anomaly over OCSVM and its varieties.