Density-Based Machine Learning Scheme for Outlier Detection in Smart Forest Fire Monitoring Sensor Cloud

Abstract

Sensor Cloud is an integration of sensor networks with cloud where sensed data is stored and processed in the cloud. The applications of sensor cloud can be seen in forest fire monitoring, healthcare system, and other Internet-of-Things systems. Outliers may present within this data due to malicious activities, low-quality sensors, or node deployment in harsh environments. Such outliers must be detected timely for effective decision making. Many clustering-based machine learning schemes for outlier detection have been devised. However, accuracy of these techniques can be further improved. This paper proposes a density-based machine learning scheme (DBS) for outlier detection which is implemented in Python and executed on the two datasets of different forest fire monitoring networks. DBS makes density-based clusters of all data points where outliers lie in low-density region. The use of a density-based model in the proposed approach improves precision, throughput, and accuracy. DBS outperforms the existing Mean Shift and K Means based clustering schemes with maximum accuracy 98.40%.