Data mining approach for earthquake data analysis

Turkey is a country which is exposed to natural features such as and primarily earthquakes then avalanches, landslides, floods and fire and will continue to remain. Therefore; together with the life and property loss caused by natural disasters economic losses should be reduced. This can be achieved with and effective study during the preparation phase of an earthquake. In this thesis the analysis of earthquake data and the evaluation with Data Mining Technique is aimed. Data Mining is the search and analysis of the significant and useful relations among huge amount of data via computer programmes. Kütahya and the region around including the circle with a radius of 250 km is selected as an application region and earthquakes' magnitude, time, depth, the existence of hot water supply, soil type which occured from 1900 until today and the distance from the centre of Kütahya was examined. For this purpose, Weka 3.6.8 Data Mining Software was used in order to make earthquake data more meaningful analyze the hidden values within and to get the opportunity to evaluate this data. Different approaches were used for classification of earthquake data and artificial neural networks' multilayer perceptron network model with the accuracy rate of 0.87 was thought to be more appropriate. In addition, more detailed analysis of earthquake data was performed with the algorithm of k-means clustering analysis.