Oktay Baykara

Artificial neural network model for earthquake prediction with radon monitoring

Apart from the linear monitoring studies concerning the relationship between radon and earthquake, an artificial neural networks (ANNs) model approach is presented starting out from non-linear changes of the eight different parameters during the earthquake occurrence. A three-layer Levenberg-Marquardt feedforward learning algorithm is used to model the earthquake prediction process in the East Anatolian Fault System (EAFS). The proposed ANN system employs individual training strategy with fixed-weight and supervised models leading to estimations. The average relative error between the magnitudes of the earthquakes acquired by ANN and measured data is about 2.3%. The relative error between the test and earthquake data varies between 0% and 12%. In addition, the factor analysis was applied on all data and the model output values to see the statistical variation. The total variance of 80.18% was explained with four factors by this analysis. Consequently, it can be concluded that ANN approach is a potential alternative to other models with complex mathematical operations.

The Statistical Analysis of the Radioactivity Concentration of the Water Data in Malatya City, Turkey

Abstract The natural radioactivity data of 306 water samples, which were obtained from various 51 sampling points of Malatya that is the eastern city of Turkey, was studied by applying the factor and cluster statistical analysis to interpretation and determination of the general radioactivity in the city. The obtained data from the statistical analysis were evaluated according to the geological formation of the state. The varimax factor rotation was applied to the data. Two factors were extracted from the radioactivity of the water supplies, tap water, and water depots data, which account for about 95.52% of the total variance. Factor 1 and factor 2 explained 48.5% and 47% of the variance, respectively.