Clustering of GPS Sensor Network Data Streams Using Self-Organizing Maps For Automatic Seismic Event Recognition
Document Type
Oral Presentation
Campus where you would like to present
SURC 140
Start Date
17-5-2012
End Date
17-5-2012
Abstract
Time series clustering of GPS sensor data in order to identify meaningful geological features and events remains a relatively unexplored field. Clustering of GPS data can potentially extract previously hidden features as well as assist in rapidly modeling geological events such as earthquakes. This is an essential task when doing things such as predicting the likelihood and location of a potential Tsunami, which can be used to mitigate disaster. In this study we will train a classifier using a self-organizing map, an artificial neural network suitable for clustering, to identify if ground movement data coming from GPS sensors indicates the occurrence of an earthquake. Associated with earthquakes are unique patterns of tiny ground shifts over time which can be used for classification. We will train the classifier to distinguish wether or not an earthquake is occurring as well as establish clustering that can be used to distinguish between different types of earthquakes. Once trained the classifier can then be used with real-time data to classify seismic phenomenon on the fly. Artificial neural networks are noise tolerant so less preprocessing needs to occur which means a rapider response. Correctly set up they can extract very subtle information, which can often be overlooked using other approaches. Finally they are adaptive and can perform under a variety of changing conditions. We will train and test the neural network using data coming from a GPS sensor network that spans the Pacific Northwest known as PANGA, which is headquarted at Central Washington University.
Recommended Citation
Smigaj, Andrew, "Clustering of GPS Sensor Network Data Streams Using Self-Organizing Maps For Automatic Seismic Event Recognition" (2012). Symposium Of University Research and Creative Expression (SOURCE). 14.
https://digitalcommons.cwu.edu/source/2012/oralpresentations/14
Additional Mentoring Department
Computer Science
Additional Mentoring Department
Geological Sciences
Clustering of GPS Sensor Network Data Streams Using Self-Organizing Maps For Automatic Seismic Event Recognition
SURC 140
Time series clustering of GPS sensor data in order to identify meaningful geological features and events remains a relatively unexplored field. Clustering of GPS data can potentially extract previously hidden features as well as assist in rapidly modeling geological events such as earthquakes. This is an essential task when doing things such as predicting the likelihood and location of a potential Tsunami, which can be used to mitigate disaster. In this study we will train a classifier using a self-organizing map, an artificial neural network suitable for clustering, to identify if ground movement data coming from GPS sensors indicates the occurrence of an earthquake. Associated with earthquakes are unique patterns of tiny ground shifts over time which can be used for classification. We will train the classifier to distinguish wether or not an earthquake is occurring as well as establish clustering that can be used to distinguish between different types of earthquakes. Once trained the classifier can then be used with real-time data to classify seismic phenomenon on the fly. Artificial neural networks are noise tolerant so less preprocessing needs to occur which means a rapider response. Correctly set up they can extract very subtle information, which can often be overlooked using other approaches. Finally they are adaptive and can perform under a variety of changing conditions. We will train and test the neural network using data coming from a GPS sensor network that spans the Pacific Northwest known as PANGA, which is headquarted at Central Washington University.
Faculty Mentor(s)
Razvan Andonie, Tim Melbourne