Maria Heinrich

[MHeini]

Author

• Name: Maria Heinrich

• Affiliation: FH Kufstein

Keywords

• earthquake prediction
• radon anomalies time series prediction
• earthquake likelihood
• natural science
• machine learning

Abstract
Earthquake prediction is currently the most important task required for probability, hazard, risk mapping, and mitigation. In the past, various traditional and machine learning models have been used for risk assessment. It is unlikely that anyone will ever be able to accurately predict earthquakes, but with the advancement of deep learning algorithms, predictions can be made more accurately and with less distance to the natural disaster. Different machine learning approaches and deep learning models based on radon anomaly detection have been compared, opening the field for further developments

..

earthquake prediction-maria-heinrich.pdf

Full PDF: DOI, [Zenodo],
Github Repo.
Github Release
Zenodo WholeTale
Zenodo Github
..

[miticdalibor]

The present literature research "Earthquake prediction with machine learning models based on peak of radon anomalies" gives very fundamental insights about earthquake prediction using radon concentration soil gases and different time series machine learning approaches. This literature research clearly states the problem statement and provides a very structured introduction for readers without any background in the earthquake research area. Explaining the forecast methodologies and data used for earthquake prediction, shows that the author gained a lot of background knowledge from different sources. The comparison of the different models and stages shows the structure of the proposed methodology. This article shows the approach using a meta learner, which can be leveraged also in other research areas using time series data and anomaly detection. For the presentation, it would be interesting to show the accuracy of the model, which uses the two-stage-approach, if this data is available. In general, this article gives fundamental insights of used machine learning models, which is a valuable contribution for data scientists and can be used as baseline for further model research in the earthquake research area.

[kayagoekan]

Order accepted for review

[renatofnc]

The paper written by Maria Heinrich analyses the possible uses of machine learning for earthquake prediction. It compares different machine learning solutions for this task, including the use of deep learning models based on radon anomaly detection. The article formulates precisely the research question, including the challenges connected to it. The author’s vivid description of how earthquakes occur from the interaction of tectonic plates facilitate the introduction into the topic. The locates the prediction problem in the area of anomaly detection regarding the element radon. Based on that, it describes concisely how the levels of the noble gas radon could be utilized in determined methods of machine learning. The author introduces the necessary methodology regarding data collection for the purpose of detecting concentration changes of radon in seismically active areas.
After setting the introductory ground, the author keeps doing a good job and exposes different forecasting methods using ensemble and individual machine learning approaches to predict the concentration of radon in the soil, under consideration of numerous environmental features. The forecasting methods are described utilizing experiments of very robust grounds, whose results are well summarized by the author.
In my final consideration I see a very concise and well written article that succeeds to introduce into the area of calamity detection following a structured path. Larger sections like the second could eventually be divided in two smaller subsections in order to increase readability, although the differentiation between experiments is still possible without further problems.

Points: (18/20)