4 results
Search Results
Now showing 1 - 4 of 4
Article Citation - WoS: 9Citation - Scopus: 8New Computational Methods for Classification Problems in the Existence of Outliers Based on Conic Quadratic Optimization(Taylor & Francis inc, 2020) Yerlikaya-Ozkurt, Fatma; Taylan, PakizeMost of the statistical research involves classification which is a procedure utilized to establish prediction models to set apart and classify new observations in the dataset from every fields of science, technology, and economics. However, these models may give misclassification results when dataset contains outliers (extreme data points). Therefore, we dealt with outliers in classification problem: firstly, by combining robustness of mean-shift outlier model and then stability of Tikhonov regularization based on continuous optimization method called Conic Quadratic Programming. These new methodologies are performed on classification dataset within the existence of outliers, and the results are compared with parametric model by using well-known performance measures.Article Citation - WoS: 9Citation - Scopus: 10Prediction of Potential Seismic Damage Using Classification and Regression Trees: a Case Study on Earthquake Damage Databases From Turkey(Springer, 2020) Yerlikaya-Ozkurt, Fatma; Askan, AysegulSeismic damage estimation is an important key ingredient of seismic loss modeling, risk mitigation and disaster management. It is a problem involving inherent uncertainties and complexities. Thus, it is important to employ robust approaches which will handle the problem accurately. In this study, classification and regression tree approach is applied on damage data sets collected from reinforced concrete frame buildings after major previous earthquakes in Turkey. Four damage states ranging from None to Severe are used, while five structural parameters are employed as damage identifiers. For validation, results of classification analyses are compared against observed damage states. Results in terms of well-known classification performance measures indicate that when the size of the database is larger, the correct classification rates are higher. Performance measures computed for Test data set indicate similar success to that of Train data set. The approach is found to be effective in classifying randomly selected damage data.Conference Object Comparison of Field Survey-Based Macroseismic Intensity Map and Numerical Macroseismic Intensity Maps Using Mars and Multiple Regression Methods for 6th of February 2023 Kahramanmaraş Earthquakes(Springer Science and Business Media Deutschland GmbH, 2024) Albayrak, Kubilay; Askan, Aysegul; Yerlikaya-Ozkurt, FatmaEarthquakes are natural events that cause damage to built environments by the instant seismic energy release. This energy is measured by instrumental devices to obtain the peak ground motion parameters such as Peak Ground Acceleration (PGA) and Peak Ground Velocity (PGV). Additional measurements based on the questionnaires after the earthquakes are required to identify the felt or macro seismic intensity levels. These measurements are crucial to identify the total effects of earthquakes over not only an area but also for the spatial distribution of ground motion parameters. For this purpose, it is important to study the multi-variable criteria correlations between ground motion parameters and Modified Mercalli Intensity (MMI) levels based on linear relationships of predictor variable couples. In this regard, the Multivariate Adaptive Regression Splines (MARS) Method and the Multiple Linear Regression Method are used. The entire dataset is composed of 69 earthquakes between 2005 and 2022 with 2171 ground motion parameters coupled with MMI levels. For MMI-based correlations, the MARS method is used to identify the non-linearities between predictor variables by piecewise linear functions, but for the Multiple Linear Regression Method, the least correlated variables of PGA-Epicentral Distance and PGV-Epicentral Distance are used to obtain the relationship between MMI and PGM parameters. The resulting equations obtained for the entire Turkiye database are performed to identify MMI maps of the 6th of February 2023, Kahramanmaras Earthquakes, and these maps are used to check the accuracy of the results by the comparison of field survey-based MMI maps. Finally, the numerical MMI maps are found to be consistent with the field survey-based MMI maps.Article Citation - WoS: 8Citation - Scopus: 8Modeling the Mood State on Thermal Sensation With a Data Mining Algorithm and Testing the Accuracy of Mood State Correction Factor(Pergamon-elsevier Science Ltd, 2025) Yerlikaya-Ozkurt, Fatma; Ozbey, Mehmet Furkan; Turhan, CihanPsychology is proven as an influencing factor on thermal sensation. On the other hand, mood state is one of the significant parameters in psychology field. To this aim, in the literature, mood state correction factor on thermal sensation (Turhan and Ozbey coefficients) is derived utilizing with data-driven black-box model. However, novel models which present analytical form of the mood state correction factor should be derived based on the several descriptive variables on thermal sensation. Moreover, the result of this factor should also be checked with analytical model results. Therefore, this study investigates the modelling of mood state correction factor with a data mining algorithm, called Multivariate Adaptive Regression Splines (MARS). Additionally, the mood state is also taken as a thermal sensation parameter besides environmental parameters in this algorithm. The same data, which are collected from a university study hall in a temperate climate zone, are used and the model results are compared with the thermal sensation results based on mood state correction factor which is driven via black-box model. The results show that coefficient of correlation "r" between the MARS and black-box model is found as 0.9426 and 0.9420 for training and testing. Hence, the mood state is also modelled via a data mining algorithm with a high accuracy, besides the black-box model.
