İlgi alanları

Sürdürülebilir Kalkınma Amaçları

• 5
• 3
• 2
• 1
• 1
• 1

• MakaleAçık Erişim2026/r/bayburt/14286
  Improvement of machine learning-based diabetes diagnosis via resampling techniques

  Yapici, Irem Senyer; Arslan, Rukiye Uzun; Engin, Mustafa Alptekin

  The objective of this study is to enhance the accuracy of diabetes diagnosis through the utilisation of machine learning techniques and resampling methods. The imbalanced nature of diabetes datasets presents a significant challenge for traditional classification algorithms, which often struggle to accurately predict results. In order to enhance the efficacy of the model, a comparative analysis was conducted to assess the performance of a range of over-sampling and under-sampling techniques, including SMOTE, ADASYN, Borderline SMOTE, SVM SMOTE, Random Under Sampler, Near Miss, One Sided Selection, Neighbourhood Cleaning Rule, Edited Nearest Neighbours, Instance Hardness Threshold, AllKNN and Tomek Links. The aforementioned techniques were then applied to the Decision Tree, Random Forest, K-Nearest Neighbours, AdaBoost, Extra Tree Classifier, and machine learning classifiers, and their performance was evaluated using the accuracy, recall, precision, F-Score, and AUC-ROC performance metrics. The SVMSMOTE resampling technique was identified as the most successful method, achieving 99.06% accuracy when used in combination with the decision tree classifier. The findings demonstrate that the incorporation of resampling techniques markedly enhances diagnostic proficiency and yields more dependable forecasts. This research makes a significant contribution to the field of medical informatics, providing a robust framework for diabetes diagnosis and offering valuable insights into the application of machine learning healthcare.

  • 
  İncele
• MakaleAçık Erişim2026Scopus SJRQ1/r/bayburt/12675
  Enhanced Absorption-Dominant EMI Shielding Performance of Pyramidal Cementitious Composites Incorporating Recycled Plastics and Magnetite Minerals for 5G Applications

  Cakir, Mehmet; Engin, Mustafa Alptekin; Camuzcuoglu, Murat

  In this study, waste polypropylene (PP) and magnetite (Fe3O4) mineral-reinforced cement-based pyramidal composite structures were designed, manufactured, and experimentally characterized to reduce electromagnetic interference (EMI) problems in the 3.3–4.9 GHz frequency band for 5G communication systems. Unlike traditional planar concrete surfaces, the aim was to minimize surface reflections and obtain an absorption-dominant shielding mechanism by providing gradient impedance matching through the pyramidal geometry. Although the use of carbon-based nanomaterials is common in the current literature, their high cost and corrosion risks limit their large-scale applications. This study involves the evaluation of waste polypropylene disposal and self-enriching magnetite mineral together. Theoretical analyses were supported by the Lichtenecker Logarithmic Mixing Rule and the Maxwell–Garnett model, and seven different mixing scenarios (S1–S7) were measured using the free-space method with a Libre vector network analyzer. Experimental results showed that the pure concrete sample exhibited predominantly reflective behaviour, with shielding performance improving significantly as the filler ratio increased. The S4 sample, containing 15% PP and 10% magnetite, offered broadband and balanced absorption performance, while the S7 sample, containing 25% PP and 25% magnetite, provided the highest shielding effectiveness with reflection below −10 dB across the entire band and transmission loss reaching −65 dB.

  • 
  • 
  İncele
• MakaleAçık Erişim2026Scopus SJRQ2/r/bayburt/11555
  Machine Learning-Based Material Thickness and Type Classification via Scattering Parameters

  Engin, Mustafa Alptekin; Cakir, Mehmet

  The precise determination of material type and thickness is of major significance in non-destructive testing, quality assurance, and materials science, as it influences the functionality, reliability, and performance of materials in engineering applications. This study proposes a methodology for the classification of material thickness and type through the analysis of scattering parameters within the 8–12 GHz frequency range. A database was created, encompassing real, imaginary, and dB values of reflection and transmission parameters for nine real-world materials with thicknesses ranging from 1 to 10 mm. This study addressed two main classification tasks, namely material thickness and material type. A variety of training–testing splits were employed in conjunction with 10-fold cross-validation to facilitate a comparison of classifier performance. In the material type classification, incorporating multiple thickness levels of each material enabled the model to distinguish materials with similar reflection characteristics more accurately, thereby enhancing discrimination performance. The results showed that Fine KNN consistently achieved 100% accuracy in thickness classification, while Quadratic SVM achieved 100% accuracy in material type classification, even when utilising only three thickness levels. Furthermore, the 90% training–10% testing split yielded the highest performance in thickness classification, whereas the optimal data split for material type classification differed across classifiers. Overall, this study demonstrates that the combination of scattering parameters and machine learning serves as a reliable and efficient approach for non-destructive material characterization.

  AtıfScopus1·WoS1·OpenAlex1

  İncele
• MakaleAçık Erişim2026Scopus SJRQ1/r/bayburt/11323
  A Single-Source Pilot Study of Machine Learning-Assisted Microwave S-Parameter Screening for Glucose Syrup and Water Adulteration in Grape Molasses

  Engin, Mustafa Alptekin; Cakir, Mehmet; Cakil, Turan

  Grape molasses, known as pekmez in Turkish, is a traditional concentrated fruit product that may be adulterated with cheaper sweeteners or water. This study evaluated broadband microwave S-parameter measurements as a rapid, non-destructive screening approach for detecting glucose syrup substitution and water dilution in grape molasses. Nine physical mixture groups were prepared, including pure grape molasses, glucose syrup–substituted mixtures at 5–30%, pure glucose syrup as an endpoint reference, and water-diluted mixtures at 10–30%. For each group, five consecutive technical measurements were recorded using a Libre vector network analyzer connected to a WR-229 waveguide-based two-port setup over 3.30–4.90 GHz. All mixtures were prepared from a single commercial grape molasses source and a single glucose syrup source; the results should therefore be interpreted as pilot-scale, single-source evidence rather than a generalizable screening method. Glucose syrup substitution produced a systematic upward shift in S11 resonance frequency. In the practical 0–30% range, S11 resonance frequency showed a strong linear relationship with glucose syrup content on the calibration data (R2 = 0.9955; inverse prediction error = 0.72 percentage points), though this reflects calibration fit rather than independently validated prediction accuracy. Water dilution produced stronger S21 attenuation. The detection principle is based on the complementary use of S11 resonance behavior for glucose syrup substitution and S21 transmission loss for water dilution. In group-blocked point-wise classification, the Ensemble Bagged Trees classifier achieved 88.79% accuracy and a macro-F1 score of 0.874. These results indicate that S11 and S21 provide complementary proof-of-concept indicators for controlled-mixture screening of grape molasses adulteration.

  • 
  İncele
• MakaleAçık Erişim2026Scopus SJRQ1/r/bayburt/11215
  Waveguide-Based Microwave S-Parameter Measurements for Machine Learning-Assisted Classification of Paprika Powder Adulterants

  Alptekin Engin, Mustafa

  Detecting food adulteration in powdered spices through visual inspection is difficult, and paprika powder is particularly susceptible to adulteration due to its fine particulate structure and intense color. This study proposes a waveguide-based microwave S-parameter measurement approach combined with machine learning for the classification of organic and mineral adulterants in paprika powder. The experimental dataset was created by intentionally adulterating a commercial paprika brand with five organic and five mineral materials in varying proportions. The S11 and S21 scattering parameters, measured in the 3.3-4.9 GHz frequency range, of these samples, comprising a total of 11 different classes including pure paprika, were used as input in the machine learning-based classification process. The proposed framework directly uses frequency and S11 and S21 responses to distinguish pure paprika from ten adulterant types without complex feature extraction. S-parameter measurements were performed iteratively for each condition using a vector network analyzer, and the average values were used in the analyses. The frequency-point-level performance of the model was examined with different training/test data splitting ratios during the classification process. Furthermore, 10-fold cross-validation was performed during the training phase, and the performance of different classifiers was evaluated comparatively using overall and balanced validation and test accuracies. Under random point-wise partitioning, Subspace KNN achieved the highest numerical performance, with a mean frequency-point-level test accuracy of 99.86±0.06% across 10 independent runs. The findings demonstrate the frequency-point-level separability of microwave S-parameter responses under the investigated experimental conditions.

  İncele
• MakaleAçık Erişim2026Scopus SJRQ2/r/bayburt/11088
  Deep Learning-Based Classification of Common Lung Sounds via Auto-Detected Respiratory Cycles

  Engin, Mustafa Alptekin; Uzun Arslan, Rukiye; Senyer Yapici, İrem; Aras, Selim; Gangal, Ali

  Chronic respiratory diseases, the third leading cause of mortality on a global scale, can be diagnosed at an early stage through non-invasive auscultation. However, effective manual differentiation of lung sounds (LSs) requires not only sharp auditory skills but also significant clinical experience. With technological advancements, artificial intelligence (AI) has demonstrated the capability to distinguish LSs with accuracy comparable to or surpassing that of human experts. This study broadly compares the methods used in AI-based LSs classification. Firstly, respiratory cycles—consisting of inhalation and exhalation parts in LSs of different lengths depending on individual variability, obtained and labelled under expert guidance—were automatically detected using a series of signal processing procedures and a database was obtained in this way. This database of common LSs was then classified using various time-frequency representations such as spectrograms, scalograms, Mel-spectrograms and gammatonegrams for comparison. The utilisation of proven, convolutional neural network (CNN)-based pre-trained models through the application of transfer learning facilitated the comparison, thereby enabling the acquisition of the features to be employed in the classification process. The performances of CNN, CNN and Long Short-Term Memory (LSTM) hybrid architecture and support vector machine methods were compared in the classification process. When the spectral structure of gammatonegrams, which capture the spectral structure of signals in the low-frequency range with high fidelity and their noise-resistant structures, is combined with a CNN architecture, the best classification accuracy of 97.3% ± 1.9 is obtained.

  • 
  AtıfScopus2·WoS2·OpenAlex2

  İncele
• MakaleAçık Erişim2026Scopus SJRQ2/r/bayburt/10926
  Enhancing salt stress tolerance in sugar beet via TiO2 nanoparticles: a machine learning approach

  Engin, Mustafa Alptekin; Sefaoglu, Firat; Gul, Volkan; Aras, Selim; Kaynar, Dilara

  In this study, the biostimulant potential of titanium dioxide (TiO2) nanoparticles was investigated to enhance sugar beet (Beta vulgaris L.) seed response to salt stress. Controlled applications were conducted at different NaCl (0–200 mM) and TiO2 (0–1,800 ppm) levels, and eight key morphological and physiological parameters were evaluated. In the first stage, the individual and interactive effects of the factors were examined using a two-way analysis of variance. Then, various machine learning-based regression models, primarily the Gradient Boosting algorithm, were used to numerically model plant responses. Based on the high-accuracy models, optimal application combinations were determined for each parameter and visualized using graphical surface analyses. Additionally, the most suitable overall TiO2 dosage for each salt level was calculated using a multi-criteria scoring system that assigned equal weight to all parameters. The results revealed that TiO2 exerts a regulatory and protective effect on plants against salt stress, with this effect varying with both dosage level and environmental stress severity. The study demonstrates that nanotechnological initiatives can be integrated into data-driven agricultural strategies to optimize stress management.

  • 
  AtıfScopus1·WoS1·OpenAlex1

  İncele
• MakaleAçık Erişim2025/r/bayburt/16001
  Speeding up the Discrete Cosine Transform Through Custom Operations and Fixed-Point Arithmetic

  Akçay, Latif; Engin, Mustafa Alptekin

  Digital signal processing applications are becoming increasingly important because modern systems work with much larger amounts of data than before. The Discrete Cosine Transform (DCT), used in almost all multimedia compression methods, creates a significant computational load especially in resource-constrained embedded systems. This study proposes four custom operations compatible with Transport-Triggered Architecture (TTA). To enhance computational efficiency and avoid floating-point overhead, fixed-point arithmetic is used. To analyse the effect of the proposed operations, different Application-Specific Instruction Set Processor (ASIP) configurations were created on a general-purpose processor architecture. Performance analyses show that speedups between 2x and 3.5x are achieved. In addition, the developed processor models have been implemented in hardware. FPGA synthesis results indicate a reasonable increase in chip area, showing that the proposed solutions could be an efficient alternative, particularly for limited-resource embedded systems.

  • 
  İncele
• MakaleAçık Erişim2024/r/bayburt/16474
  Depresyonda motor aktivitenin makine öğrenmesi ile değerlendirilmesi

  Aras, Selim; Aras, Neriman; Engin, Mustafa Alptekin

  Psikiyatrik hastalıkların neredeyse tümünde olduğu gibi depresyonun da klinik olarak değerlendirilmesi gözleme ve subjektif hasta şikâyetlerine dayanmaktadır. Psikomotor retardasyon (gerileme) depresyonun önde gelen semptomlarından biridir ve bunun göstergesi olarak depresyonlu hastalarda fiziksel aktivite azalır. Bu çalışmada, depresyonu olan ve olmayan bireylerin günlük fiziksel aktivite verileri ile oluşturulmuş bir veri setini referans olarak kullanarak, depresyon tanısı için makine öğrenimi temelli objektif bir tanı destekleyici yöntem geliştirmek amaçlanmıştır. Geniş bir öznitelik araştırması yapıldıktan sonra, Fisher Öznitelik Seçimi ile en iyi dört öznitelik belirlenmiş ve Toplu Torbalı Ağaç yöntemini kullanarak 0,88 doğruluk ile referans çalışmasından daha iyi bir sınıflandırma sonucu elde edilmiştir. Ayrıca, referans çalışma ile karşılaştırmak için sınırlandırılan dört öznitelikten daha fazlası seçildiğinde doğruluğun 0,90’nın üzerine çıktığı belirlenmiştir. Böylece, fiziksel aktivite verilerini kullanarak geliştirilen makine öğrenimi temelli yöntemle depresyonu olan ve olmayan bireyleri yüksek doğruluk payı ile ayırt edilmesi başarılmıştır. Bu çalışma, aktivite verilerinin depresyonda tanı destekleyici bir araç olarak kullanılabileceğine dair umut verici sonuçlar ortaya koymuştur. Elde edilen sonuçlar, fiziksel aktivite gibi farklı biyobelirteçlerin de makine öğrenimi ile birlikte kullanıldığında, psikiyatrik değerlendirmedeki objektif tanı destekleyici kriterlerin eksikliğini giderebilecek potansiyele sahip olduğunu göstermektedir.

  • 
  İncele
• MakaleAçık Erişim2024/r/bayburt/16390
  Parkinson’s Disease Detection Via Machine Learning Using Data Splitting and Validation Methods

  Engin, Mustafa Alptekin

  Parkinson’s disease (PD), a neurological disorder, negatively affects the lives of patients and their caregivers. PD, which is very difficult to diagnose early by examining the clinical characteristics of the person, can be diagnosed using voice recordings. However, the inconsistent performance results of the models obtained from the evaluation of voice recordings through machine learning techniques limit the usability of these models to aid in diagnosing physicians. This study used a database of 195 voice data obtained from 31 individuals, 23 of whom have PD. The classification of the voices as healthy or patient was based on the 22 features in the database. The split ratios 90/10, 80/20, 70/30, 50/50 and 30/70 were used to select the training and test phase data, respectively. In addition, each split ratio was evaluated using 10-fold cross-validation, 5-fold cross-validation, holdout validation and resubstitution validation methods in the training phase, which is the initial process that will directly affect the other classification procedures. In addition, the classification process was performed using quadratic discriminant analysis, support vector machine, ensemble bagged tree, k-nearest neighbours and neural network classifiers. All procedures were repeated 10 times to ensure consistency of results and randomisation of split ratios. As a result, the k-nearest neighbours classifier with 80/20 splitting ratio and 10-fold cross-validation was determined to be the most successful among the compared methods with 95.64±3.21% accuracy. Therefore, it can be seen that much more successful results can be obtained by analysing only the effects of the existing parameters of the classifiers.

  • 
  İncele