WoS İndeksli Yayınlar Koleksiyonu / WoS Indexed Publications Collection
Permanent URI for this collectionhttps://hdl.handle.net/11147/7150
Browse
8 results
Search Results
Article Citation - WoS: 15Citation - Scopus: 16A Machine Learning Ensemble Approach for Predicting Solar-Sensitive Hybrid Photocatalysts on Hydrogen Evolution(IOP Publishing, 2024) Bakır, Rezan; Orak, Ceren; Yuksel, AsliHydrogen, as the lightest and most abundant element in the universe, has emerged as a pivotal player in the quest for sustainable energy solutions. Its remarkable properties, such as high energy density and zero emissions upon combustion, make it a promising candidate for addressing the pressing challenges of climate change and transitioning towards a clean and renewable energy future. In an effort to improve efficiency and reduce experimental costs, we adopted machine learning techniques in this study. Our focus turned to predictive analyses of hydrogen evolution values using three photocatalysts, namely, graphene-supported LaFeO3 (GLFO), graphene-supported LaRuO3 (GLRO), and graphene-supported BiFeO3 (GBFO), examining their correlation with varying levels of pH, catalyst amount, and H2O2 concentration. To achieve this, a diverse range of machine learning models are used, including Random Forest (RF), Decision Tree (DT), Support Vector Machine (SVM), XGBoost, Gradient Boosting, and AdaBoost-each bringing its strengths to the predictive modeling arena. An important step involved combining the most effective models-Random Forests, Gradient Boosting, and XGBoost-into an ensemble model. This collaborative approach aimed to leverage their collective strengths and improve overall predictability. The ensemble model emerged as a powerful tool for understanding photocatalytic hydrogen evolution. Standard metrics were employed to assess the performance of our ensemble prediction model, encompassing R squared, Root Mean Squared Error (RMSE), Mean Squared Error (MSE), and Mean Absolute Error (MAE). The yielded results showcase exceptional accuracy, with R squared values of 96.9%, 99.3%, and 98% for GLFO, GBFO, and GLRO, respectively. Moreover, our model demonstrates minimal error rates across all metrics, underscoring its robust predictive capabilities and highlighting its efficacy in accurately forecasting the intricate relationships between GLFO, GBFO, and GLRO values and their influencing factors.Article Citation - WoS: 3Citation - Scopus: 3Ensemble and Optimized Hybrid Algorithms Through Runge Kutta Optimizer for Sewer Sediment Transport Modeling Using a Data Pre-Processing Approach(Elsevier, 2023) Safari, Mir Jafar Sadegh; Gül, Enes; Dursun, Ömer Faruk; Tayfur, GökmenUncontrolled sediment deposition in drainage and sewer systems raises unexpected maintenance expenditures. To this end, implementation of an accurate model relying on effective parameters involved is a reliable benchmark. In this study, three machine learning techniques, namely extreme learning machine (ELM), multilayer perceptron neural network (MLPNN), and M5P model tree (M5PMT); and three optimization approaches of Runge Kutta (RUN), genetic algorithm (GA), and particle swarm optimization (PSO) are applied for modeling. The optimization and ensemble hybridization approaches are applied in the modeling procedure. For the case of hybrid optimized models, the ELM and MLPNN models are hybridized with RUN, GA, and PSO algorithms to develop six hybrid models of ELM-RUN, ELM-GA, ELM-PSO, MLPNN-RUN, MLPNN-GA, and MLPNN-PSO. Ensemble hybrid models are developed through coupling the ELM and MLPNN models with the M5PMT algorithm. The data pre-processing approach is applied to find the best randomness characteristic of the utilized data. Results illustrate that the RUN-based hybrid models outperform the GA- and PSO-based counterparts. Although the MLPNN-RUN and MLPNN-M5PMT hybrid models generate better results than their alternatives, MLPNN-M5PMT slightly outperforms MLPNN-RUN model with a coefficient of determination of 0.84 and a root mean square error of 0.88. The current study shows the superiority of the ensemble-based approach to the optimization techniques. Further investigation is needed by considering alternative optimization techniques to enhance sediment transport modeling. © 2023 International Research and Training Centre on Erosion and Sedimentation/the World Association for Sedimentation and Erosion ResearchArticle Citation - WoS: 13Citation - Scopus: 17Machine Learning-Assisted Prediction of the Toxicity of Silver Nanoparticles: a Meta-Analysis(Springer, 2023) Bilgi, Eyüp; Öksel Karakuş, CeydaSilver nanoparticles are likely to be more dangerous than other forms of silver due to the intracellular release of silver ions upon dissolution and the formation of mixed ion-containing complexes. Such concerns have resulted in an ever-growing pile of scientific evaluations addressing the safety aspects of nanosilver with widely varying methodological approaches. The substantial differences in the conduct/design of nanotoxicity screening have led to the generation of conflicting findings that may be accurate in their narrative but fail to provide a complete picture. One strategy to maximize the use of individual risk assessments with potentially biased estimates of toxicological effects is to homogenize results across several studies and to increase the generalizability and human relevance of their findings. Here, we collected a large pool of data (n=162 independent studies) on the cytotoxicity of nanosilver and unrevealed potential triggers of toxicity. Two different machine learning approaches, decision tree (DT) and artificial neural network (ANN), were primarily employed to develop models that can predict the cytotoxic potential of nanosilver based on material- and assay-related parameters. Other machine learning algorithms (logistic regression, Gaussian Naive Bayes, k-nearest neighbor, and random forest classifiers) were also applied. Among several attributes compared, exposure concentration, duration, zeta potential, particle size, and coating were found to have the most substantial impact on nanotoxicity, with biomolecule- and microorganism-assisted surface modifications having the most beneficial and detrimental effects on cell survival, respectively. Such machine learning-assisted efforts are critical to developing commercially viable and safe nanosilver-containing products in the ever-expanding nanobiomaterial market.Article Performance and Accuracy Predictions of Approximation Methods for Shortest-Path Algorithms on Gpus(Elsevier, 2022) Aktılav, Busenur; Öz, IşılApproximate computing techniques, where less-than-perfect solutions are acceptable, present performance-accuracy trade-offs by performing inexact computations. Moreover, heterogeneous architectures, a combination of miscellaneous compute units, offer high performance as well as energy efficiency. Graph algorithms utilize the parallel computation units of heterogeneous GPU architectures as well as performance improvements offered by approximation methods. Since different approximations yield different speedup and accuracy loss for the target execution, it becomes impractical to test all methods with various parameters. In this work, we perform approximate computations for the three shortest-path graph algorithms and propose a machine learning framework to predict the impact of the approximations on program performance and output accuracy. We evaluate random predictions for both synthetic and real road-network graphs, and predictions of the large graph cases from small graph instances. We achieve less than 5% prediction error rates for speedup and inaccuracy values.Article Citation - WoS: 27Citation - Scopus: 34Multi-Zone Optimisation of High-Rise Buildings Using Artificial Intelligence for Sustainable Metropolises. Part 2: Optimisation Problems, Algorithms, Results, and Method Validation(Pergamon-Elsevier Science LTD, 2021) Ekici, Berk; Kazanasmaz, Zehra Tuğçe; Turrin, Michela; Taşgetiren, M. Fatih; Sarıyıldız, I. SevilHigh-rise building optimisation is becoming increasingly relevant owing to global population growth and urbanisation trends. Previous studies have demonstrated the potential of high-rise optimisation but have been focused on the use of the parameters of single floors for the entire design; thus, the differences related to the impact of the dense surroundings are not taken into consideration. Part 1 of this study presents a multi-zone optimisation (MUZO) methodology and surrogate models (SMs), which provide a swift and accurate prediction for the entire building design; hence, the SMs can be used for optimisation processes. Owing to the high number of parameters involved in the design process, the optimisation task remains challenging. This paper presents how MUZO can cope with an enormous number of parameters to optimise the entire design of high-rise buildings using three algorithms with an adaptive penalty function. Two design scenarios are considered for quad-grid and diagrid shading devices, glazing type, and building-shape parameters using the setup, and the SMs developed in part 1. The optimisation part of the MUZO methodology reported satisfactory results for spatial daylight autonomy and annual sunlight exposure by meeting the Leadership in Energy and Environmental Design standards in 19 of 20 optimisation problems. To validate the impact of the methodology, optimised designs were compared with 8748 and 5832 typical quad-grid and diagrid scenarios, respectively, using the same design parameters for all floor levels. The findings indicate that the MUZO methodology provides significant improvements in the optimisation of high-rise buildings in dense urban areas.Article Citation - WoS: 10Citation - Scopus: 13Overcoming Roadblocks in Computational Roadmaps To the Future for Safe Nanotechnology(IOP Publishing, 2021) Öksel Karakuş, Ceyda; Winkler, David A.The rapid rise of nanotechnology has resulted in a parallel rise in the number of products containing nanomaterials. The unusual properties that nano forms of materials exhibit relative to the bulk has driven intense research interest and relatively rapid adoption by industry. Regulatory agencies are charged with protecting workers, the public, and the environment from any adverse effects of nanomaterials that may also arise because of these novel physical and chemical properties. They need data and models that allow them to flag nanomaterials that may be of concern, while balancing potential stifling of commercial innovation. Roadmaps for the future of safe nanotechnology were defined more than a decade ago, but many roadblocks identified in these studies remain. Here, we discuss the roadblocks that are still hindering the effective application of informatics and predictive computational nanotoxicology methods from providing more effective guidance to nanomaterials regulatory agencies and safe-by-design rationale for industry. We describe how developments in high throughput synthesis, characterization, and biological assessment of nanomaterials will overcome many of these roadblocks, allowing a clearly defined roadmap for computational design of effective but safe-by-design nanomaterials to be realized.Article Citation - WoS: 38Citation - Scopus: 50Multi-Zone Optimisation of High-Rise Buildings Using Artificial Intelligence for Sustainable Metropolises. Part 1: Background, Methodology, Setup, and Machine Learning Results(Elsevier Ltd., 2021) Ekici, Berk; Kazanasmaz, Zehra Tuğçe; Turrin, Michela; Taşgetiren, M. Fatih; Sarıyıldız, I. SevilDesigning high-rise buildings is one of the complex tasks of architecture because it involves interdisciplinary performance aspects in the conceptual phase. The necessity for sustainable high-rise buildings has increased owing to the demand for metropolises based on population growth and urbanisation trends. Although artificial intelligence (AI) techniques support swift decision-making when addressing multiple performance aspects related to sustainable buildings, previous studies only examined single floors because modelling and optimising the entire building requires extensive computational time. However, different floor levels require various design decisions because of the performance variances between the ground and sky levels of high-rises in dense urban districts. This paper presents a multi-zone optimisation (MUZO) methodology to support decision-making for an entire high-rise building considering multiple floor levels and performance aspects. The proposed methodology includes parametric modelling and simulations of high-rise buildings, as well as machine learning and optimisation as AI methods. The specific setup focuses on the quad-grid and diagrid shading devices using two daylight metrics of LEED: spatial daylight autonomy and annual sunlight exposure. The parametric model generated samples to develop surrogate models using an artificial neural network. The results of 40 surrogate models indicated that the machine learning part of the MUZO methodology can report very high prediction accuracies for 31 models and high accuracies for six quad-grid and three diagrid models. The findings indicate that the MUZO can be an important part of designing high-rises in metropolises while predicting multiple performance aspects related to sustainable buildings during the conceptual design phase. © 2021 The Author(s)Article Citation - WoS: 14Citation - Scopus: 12The Impact of Feature Selection on One and Two-Class Classification Performance for Plant Micrornas(PeerJ Inc., 2016) Khalifa, Waleed; Yousef, Malik; Saçar Demirci, Müşerref Duygu; Allmer, JensMicroRNAs (miRNAs) are short nucleotide sequences that form a typical hairpin structure which is recognized by a complex enzyme machinery. It ultimately leads to the incorporation of 18-24 nt long mature miRNAs into RISC where they act as recognition keys to aid in regulation of target mRNAs. It is involved to determine miRNAs experimentally and, therefore, machine learning is used to complement such endeavors. The success of machine learning mostly depends on proper input data and appropriate features for parameterization of the data. Although, in general, two-class classification (TCC) is used in the field; because negative examples are hard to come by, one-class classification (OCC) has been tried for pre-miRNA detection. Since both positive and negative examples are currently somewhat limited, feature selection can prove to be vital for furthering the field of pre-miRNA detection. In this study, we compare the performance of OCC and TCC using eight feature selection methods and seven different plant species providing positive pre-miRNA examples. Feature selection was very successful for OCC where the best feature selection method achieved an average accuracy of 95.6%, thereby being ~29% better than the worst method which achieved 66.9% accuracy. While the performance is comparable to TCC, which performs up to 3% better than OCC, TCC is much less affected by feature selection and its largest performance gap is ~13% which only occurs for two of the feature selection methodologies. We conclude that feature selection is crucially important for OCC and that it can perform on par with TCC given the proper set of features.