EXPLORING CLUSTERING ALGORITHMS FOR THE ANALYSIS OF HIGH-RATE POSITIONING DATA FROM GNSS SENSORS ON OL DOINYO LENGAI

Clustering algorithms can be used to analyze large amounts of high-rate positioning data that would otherwise be very taxing to interpret. Such algorithms can be applied to analyzing noisy high-rate positioning data from Global Navigation Satellite System (GNSS) sensors that monitor an active volcano called Ol Doinyo Lengai in Tanzania. Previous work has focused on using K-means and Gaussian mixtures clustering algorithms to analyze the GNSS positioning data for hazardous signals in the height component. This research explores other clustering algorithms such as the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) and Balanced Iterative Reducing and Clustering using Hierarchies (BIRCH) algorithms. We use these clustering algorithms to find outliers in the height component of the positioning data. The ability of the clustering algorithms to handle the datasets are measured using qualitative metrics such as visualizing the data to observe how many clusters formed from the datapoints given in the dataset. Other more quantitative metrics are also used such as Silhouette Scores, which measure the Euclidean distance between data points in their assigned clusters. Further research could yield results in how to use DBSCAN and BIRCH more efficiently and how other clustering algorithms could work with the positioning data towards rapid automation of detecting hazardous volcanic signals in GNSS positioning data.