Comparison Of Model-Based Clustering to Other Clustering Methods: an Example on Meteorological Data

Abstract

Using a methodology comprised of model-based clustering and panel data analysis, we have tried to draw conclusions from a real-life meteorological data based on philosophy of science. In this study, we used model-based clustering on a real-life meteorological data consisting of yearly observations of annual mean temperatures gathered from the 58 stations in regions of Turkey and compared the results to the results of other agglomerative clustering methods that were derived upon the results of an earlier study on the aforementioned real-life meteorological data. We then configured the clusters as separate dummy variables the effects of which was put to the test by longitudinal data analysis. The agglomerative clustering results were more successful according to the between R^2 obtained from longitudinal data analysis compared to the earlier study but the best results were obtained from the model-based clustering of the aforementioned real-life meteorological data. The comparative clustering analysis demonstrated that the climate change occurs differently across regions of Turkey.

Keywords

• Longitudinal Data
• Model-Based Clustering
• Climate Change
• Philosophy of Science

References

1. Bhaumik, D., Sengupta, D. (2020), Estimating Historic Movement of a Climatological Variable From a Pair of Misaligned Functional Data Sets, Environmental and Ecological Statistics, 27, 729-751
2. Çelebioğlu, S. (2018), On Some Climatic Scenarios For Turkey From The Perspective of Changes in the Annual Mean Temperatures via Aggregation by Steady-State Distribution, International Journal of Environment and Geoinformatics,5(2), 197-217.
3. Dexen, D.Z.Xi, Dean, C.B., Taylor, S.W.(2019), Modeling The Duration and Size of Extended Attack WildFires As Dependent Outcomes, Environmetrics, 31(5).
4. McNicholas, P.D. (2017), Mixture Model-Based Classification, CRC press, New York.