Multivariate weather anomaly detection using DBSCAN clustering algorithm
Supri Wibisono, Muchamad Taufiq Anwar, Arif Supriyanto, Imam Husni Al Amin
Abstract
Abstract Weather is highly influential for human life. Weather anomalies describe conditions that are out of the ordinary and need special attention because they can affect various aspects of human life both socially and economically and also can cause natural disasters. Anomaly detection aims to get rid of unwanted data (noise, erroneous data, or unwanted data) or to study the anomaly phenomenon itself (unusual but interesting). In the absence of an anomaly-labeled dataset, an unsupervised Machine Learning approach can be utilized to detect or label the anomalous data. This research uses the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm to separate between normal and anomalous weather data by considering multiple weather variables. Then, PCA is used to visualize the clusters. The experimental result had demonstrated that DBSCAN is capable of identifying peculiar data points that are deviating from the ‘normal’ data distribution.