Talas Fikret Kurnaz Talas Fikret Kurnaz TEKNİK BİLİMLER MESLEK YÜKSEKOKULU ULAŞTIRMA HİZMETLERİ BÖLÜMÜ ULAŞTIRMA VE TRAFİK HİZMETLERİ PR.
Araştırmacının yayınlarını daha detaylı bir şekilde aramak veya filtrelemek için buraya tıklayabilirsiniz.

Estimation of soil liquefaction using artificial intelligence techniques: an extended comparison between machine and deep learning approaches

Şehmusoğlu, Eyyüp Hakan | Kurnaz, Talas Fikret | Erden, Caner

This study investigates the effectiveness of various deep learning (DL) algorithms in predicting soil liquefaction susceptibility. We explore a spectrum of algorithms, including machine learning models such as Support Vector Machines (SVMs), K-Nearest Neighbors (KNN), and Logistic Regression (LR), alongside DL architectures like Convolutional Neural Networks (CNNs), Long Short-Term Memory networks (LSTMs), Bidirectional LSTMs (BiLSTMs), and Gated Recurrent Units (GRUs). The performance of these algorithms is assessed using comprehensive metrics, including accuracy, precision, recall, F1-score, receiver operating characteristic (ROC) curve analysis, and area under the curve (AUC). Cross-entropy loss is employed as the loss function during model training to optimize the differentiation between liquefiable and nonliquefiable soil samples. Our findings reveal that the GRU model achieved the highest overall accuracy of 0.98, followed by the BiLSTM model with an accuracy of 0.95. Notably, the BiLSTM model excelled in precision for class 1, attaining a score of 0.96 on the test dataset. These results underscore the potential of both GRU and BiLSTM models in predicting soil liquefaction susceptibility, with the BiLSTM model’s simpler architecture proving particularly effective ...

An integrated SHAP-MCDM approach for slope stability prediction based on machine learning algorithms

Demir, Alparslan Serhat | Dağdeviren, Uğur | Kurnaz, Talas Fikret | Erden, Caner | Kökçam, Abdullah Hulusi

In machine learning (ML)-based slope stability prediction studies, feature importance results often vary across different algorithms, leading to inconsistent interpretations. This issue arises because the importance of features differs depending on the algorithm applied within the same study. To address this challenge, this study proposes a novel methodology for obtaining a final, unified ranking of features by combining the feature importance rankings of various ML algorithms using a Multi-Criteria Decision-Making (MCDM) technique. This approach ensures a consistent and reliable feature ranking derived from the results of successful ML models. Furthermore, the study demonstrates how performance indicators of ML algorithms can be translated into criterion weights within the MCDM framework. Hyperparameter optimization was applied to the ML models, achieving accuracy rates between 70% and 92.5%. Successful algorithms were analyzed using SHapley Additive Explanations (SHAP) to evaluate feature importance, and the results were integrated into the proposed SHAP-MCDM methodology. The MULTIMOORA method, a well-established MCDM technique, was employed to combine the SHAP rankings. The results confirmed that a final feature ranking could be derived by merging different SHAP ...

Explainable AI using ensemble machine learning with integrated SHapley additive explanations (SHAP)-Borda approach for estimation of the safety factor against soil liquefaction

Dağdeviren, Uğur | Demir, Alparslan Serhat | Erden, Caner | Kökçam, Abdullah Hulusi | Kurnaz, Talas Fikret

In most of the studies on soil liquefaction prediction based on Machine Learning (ML), the models presented are presented in a closed box structure. In the studies where the effect of the features on the model performance is analyzed with Interpretability methods, it is seen that the order of effect of the features changes for each ML algorithm. This situation makes the results of the studies conducted on the same subject inconsistent. In this study, we propose an integrated SHapley Additive exPlanations (SHAP)-Borda approach to overcome this problem. With this study, we provide decision makers with ease in explaining ML models by combining SHAP analysis results with the Borda method for the first time. In the study, ensemble ML algorithms were used for soil liquefaction prediction using data collected from the literature. The performances of the model predictions obtained by hyper parameterization were compared, and correlation results ranging from 0.91 to 0.93 were obtained. Ensemble ML algorithms that were found to be successful as a result of evaluating other performance criteria were analyzed with the SHAP-Borda approach in the study. It has been observed that with the proposed SHAP-Borda approach, the interpretability results of different ML algorithms can be brou...