The Mt. Etna data mining software

Abstract

Mt. Etna is permanently active requiring a continuous data acquisition a multidisciplinary monitoring system where huge data masses accumulate and pose severe difficulties of interpretation. Therefore the INGV staff has developed a number of software tools for data mining, aiming at identifying structures in the data which can be related to the volcanic activity and furnish criteria for the definition of alert systems. We tackle the problem by applying methods of clustering and classification. We identify data groups by defining a measure of similarity or distance. Data groups may assume various shapes, once forming convex clouds once complex concave bodies. The tool “KKAanalysis” is a basket of clustering methods and forms the backbone of the tremor-based automatic alarm system of INGV-OE. It exploits both SOM and Fuzzy Clustering. Besides seismic data the concept has been applied to petrochemic data as well as in a combined analysis of gas-emission data and seismic data. The software “DBSCAN” focuses on density-based clustering that allows discovering clusters with arbitrary shape. Here, clusters are defined as dense regions of objects in the data space separated by regions of low density. In DBSCAN a cluster grows guaranteeing that the density within a group of objects exceeds some threshold. In the context of volcano monitoring the method is particularly promising in the recognition of ash particles as they have a rather irregular shape. The “MOTIF” software allows identifying typical wave forms in time series. It overcomes shortages of methods like cross- correlation, which entail a high computational effort. MOTIF on the other hand can recognize non-similarity of two patterns on a small number of data points without going through the whole length of the data vectors. The development includes modules for feature extraction and post-processing verifying the validity of the results obtained by the classifiers.