Yazar "Acmali, Suheda Semih" seçeneğine göre listele

Listeleniyor 1 - 2 / 2
  • Küçük Resim Yok
    Öğe
    Developing comprehensive geocomputation tools for landslide susceptibility mapping: LSM tool pack
    (Pergamon-Elsevier Science Ltd, 2020) Sahin, Emrehan Kutlug; Colkesen, Ismail; Acmali, Suheda Semih; Akgun, Aykut; Aydinoglu, Arif Cagdas
    The primary aim of this research paper is to develop an easy-to-use tool package called Landslide Susceptibility Mapping Tool Pack (LSM Tool Pack) for producing landslide susceptibility maps based on integrating R with ArcMap Software. The proposed tool contains 5 main modules namely: (1) Data Preparation (DP), (2) Feature (Factor) Selection (FS), (3) Logistic Regression (LR), (4) Random Forest (RF) and (5) Performance Evaluation (PE). The FS module brings a novel approach to determine the best factor subset in the production of landslide susceptibility maps. The feature ranking values of factors were calculated by several feature ranging methods (i. e. chi-square, information gain, rank correlation, and random forest feature importance). The logistic regression method was used at the model prediction stage for each feature ranking and different models were produced for each ranking result. And, in the last step of the FS analysis, tests of statistical significance (i.e. Wilcoxon signedrank test, F- Test, Kolmogorov Smirnov test, and One-Sample t-test) were used to determine the significance of the difference between models. As a result, the best factor sets determined by the FS module were used as input factors in the LR Module and the RF Module to produce LSMs. Also, users can calculate the performance metric of landslide susceptibility maps by several performance metrics (overall accuracy, Area under the ROC Curve (AUC) value, kappa, Fl score, and more) with additional integrated the PE Module in ArcMap Software. The LSM Tool Pack is applied to the Sinop province of the Black Sea region of Turkey. Considered the FS module, Case 1 was selected as the experimental dataset for this present study. In the selected Case 1, feature ranking method and statistically significant analysis were determined by Chi-square and F-Test, respectively. As a result, Model-12, which is contained 12 landslide causative factors, was determined as the optimum subset. According to the results obtained by the accuracy assessment process, the RF model showed the best prediction performance with an AUC value of 0.8898. On the other hand, the calculated AUC value was 0.8119 for the LR model. The experimental results (using with dataset in actual study area) confirm the ability of the proposed feature selection approach in the landslide susceptibility mapping process.
  • Küçük Resim Yok
    Öğe
    Green AI-Driven Concept for the Development of Cost-Effective and Energy-Efficient Deep Learning Method: Application in the Detection of Eimeria Parasites as a Case Study
    (Wiley, 2024) Acmali, Suheda Semih; Ortakci, Yasin; Seker, Huseyin
    Although large-scale pretrained convolutinal neural networks (CNN) models have shown impressive transfer learning capabilities, they come with drawbacks such as high energy consumption and computational cost due to their potential redundant parameters. This study presents an innovative weight-level pruning technique that mitigates the challenges of overparameterization, and subsequently minimizes the electricity usage of such large deep learning models. The method focuses on removing redundant parameters while upholding model accuracy. This methodology is applied to classify Eimeria species parasites from fowls and rabbits. By leveraging a set of 27 pretrained CNN models with a number of parameters between 3.0M and 118.5M, the framework has identified a 4.8M-parameter model with the highest accuracy for both animals. The model is then subjected to a systematic pruning process, resulting in an 8% reduction in parameters and a 421M reduction in floating point operations while maintaining the same classification accuracy for both fowls and rabbits. Furthermore, unlike the existing literature where two separate models are created for rabbits and fowls, this article presents a combined model with 17 classes. This approach has resulted in a CNN model with nearly 50% reduced parameter size while retaining the same accuracy of over 90%.