Yazar "Acikkar, Mustafa" seçeneğine göre listele
Listeleniyor 1 - 5 / 5
Sayfa Başına Sonuç
Sıralama seçenekleri
Öğe A novel hybrid PSO- and GS-based hyperparameter optimization algorithm for support vector regression(Springer London Ltd, 2023) Acikkar, Mustafa; Altunkol, YunusHyperparameter optimization is vital in improving the prediction accuracy of support vector regression (SVR), as in all machine learning algorithms. This study introduces a new hybrid optimization algorithm, namely PSOGS, which consolidates two strong and widely used algorithms, particle swarm optimization (PSO) and grid search (GS). This hybrid algorithm was experimented on five benchmark datasets. The speed and the prediction accuracy of PSOGS-optimized SVR models (PSOGS-SVR) were compared to those of its constituent algorithms (PSO and GS) and another hybrid optimization algorithm (PSOGSA) that combines PSO and gravitational search algorithm (GSA). The prediction accuracies were evaluated and compared in terms of root mean square error and mean absolute percentage error. For the sake of reliability, the results of the experiments were obtained by performing 10-fold cross-validation on 30 runs. The results showed that PSOGS-SVR yields prediction accuracy comparable to GS-SVR, performs much faster than GS-SVR, and provides better results with less execution time than PSO-SVR. Besides, PSOGS-SVR presents more effective results than PSOGSA-SVR in terms of both prediction accuracy and execution time. As a result, this study proved that PSOGS is a fast, stable, efficient, and reliable algorithm for optimizing hyperparameters of SVR.Öğe An improved KNN classifier based on a novel weighted voting function and adaptive k-value selection(Springer London Ltd, 2024) Acikkar, Mustafa; Tokgoz, SelcukThis paper presents a modified KNN classifier (HMAKNN) based on the harmonic mean of the vote and average distance of the neighbors of each class label combined with adaptive k-value selection. Within the scope of this study, two different versions of HMAKNN, regular and weighted, HMAKNN(R) and HMAKNN(W), were developed depending on whether there is a weighting mechanism or not. These proposed HMAKNN classifiers were tested eight syntetic and twenty-six real benchmark data sets. In order to reveal the effectiveness and the performance of the proposed methods on classification, they were compared with its constituent KNN and four other well-known distance-weighted KNN methods. Unlike other weighting methods, both HMAKNN classifiers use the synergy between majority voting and average distance together, along with the ability to adaptively adjust the k-value, helping to significantly improve classification accuracy. The results on twenty-six real benchmark data sets suggest that both HMAKNN methods produce more accurate results in terms of average ACC and FScore metrics and statistically outperform all competing methods.Öğe Fast grid search: A grid search-inspired algorithm for optimizing hyperparameters of support vector regression(Tubitak Scientific & Technological Research Council Turkey, 2024) Acikkar, MustafaThis study presents a fast hyperparameter optimization algorithm based on the benefits and shortcomings of the standard grid search (GS) algorithm for support vector regression (SVR). This presented GS -inspired algorithm, called fast grid search (FGS), was tested on benchmark datasets, and the impact of FGS on prediction accuracy was primarily compared with the GS algorithm on which it is based. To validate the efficacy of the proposed algorithm and conduct a comprehensive comparison, two additional hyperparameter optimization techniques, namely particle swarm optimization and Bayesian optimization, were also employed in the development of models on the given datasets. The evaluation of the models' predictive performance was conducted by assessing root mean square error, mean absolute error, and mean absolute percentage error. In addition to these metrics, the number of evaluated submodels and the time required for optimization were used as determinative performance measures of the presented models. Experimental results proved that the FGS-optimized SVR models yield precise performance, supporting the reliability, validity, and applicability of the FGS algorithm. As a result, the FGS algorithm can be offered as a faster alternative in optimizing the hyperparameters of SVR in terms of execution time.Öğe Prediction of gross calorific value of coal based on proximate analysis using multiple linear regression and artificial neural networks(Tubitak Scientific & Technological Research Council Turkey, 2018) Acikkar, Mustafa; Sivrikaya, OsmanGross calorific value (GCV) of coal was predicted by using as-received basis proximate analysis data. Two main objectives of the study were to develop prediction models for GCV using proximate analysis variables and to reveal the distinct predictors of GCV. Multiple linear regression (MLR) and artifcial neural network (ANN) (multilayer perceptron MLP, general regression neural network GRNN, and radial basis function neural network RBFNN) methods were applied to the developed 11 models created by different combinations of the predictor variables. By conducting 10fold cross-validation, the prediction accuracy of the models has been tested by using R-2, RMSE, MAE, and MAPE. In this study, for the first time in the literature, for a single dataset, maximum number of coal samples were utilized and GRNN and RBFNN methods were used in GCV prediction based on proximate analysis. The results showed that moisture and ash are the most discriminative predictors of GCV and the developed RBFNN-based models produce high performance for GCV prediction. Additionally, performances of the regression methods, from the best to the worst, were RBFNN, GRNN, MLP, and MLR.Öğe Prediction of landslide tsunami run-up on a plane beach through feature selected MLP-based model(Elsevier, 2024) Aydin, Baran; Yaguzluk, Sava; Acikkar, MustafaWe proposed new prediction models based on multilayer perceptron (MLP) which successfully predict the maximum run-up of landslide -generated tsunami waves and assess the role of parameters affecting it. The input is approximately 55,0 0 0 rows of data generated through an analytical solution employing slide's cross section, initial submergence, vertical thickness, horizontal length, beach slope angle and the maximum run-up itself, along with its occurrence time. The parameters are first ranked through a feature selection algorithm and six models are constructed for a 9,0 0 0 -row randomly sampled dataset. These MLP-based models led predictions with a minimum Mean Absolute Percentage Error of 1.1% and revealed that vertical slide thickness has the largest impact on the maximum tsunami run-up, whereas beach slope angle has minimal effect. Com parison with existing literature showed the reliability and applicability of the offered models. The methodology introduced here can be suggested as fast and flexible method for prediction of landslide -induced tsunami run-up. (c) 2022 Shanghai Jiaotong University. Published by Elsevier B.V. This is an open access article under the CC BY -NC -ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ )