Browsing by Author "Mohammed, Emad"
Now showing 1 - 4 of 4
Results Per Page
Sort Options
- ItemOpen AccessAnalysis of Alternatives and Performance Evaluation Using a New OWA Operator based on the Laplace Distribution(2016) Mohammed, Emad; Far, Behrouz; Naugler, Christopher; Moussavi, Mahmood; Eberlien, Armin; Zareipour, Hamid; Sadaoui, SamiraAnalysis of Alternatives (AoA) is an assessment of operational effectiveness, appropriateness, cost, and risk associated with alternative solutions to specific problem requirements. Decision makers can utilize the AoA outcome to support their informed decisions that favor a specific alternative. Multiple criteria decision-making (MCDM) denotes the act of choosing, implementing, and applying a specific course of action to solve problems based on the AoA result of the multiple criteria alternatives. An intrinsic characteristic of the criteria is their conflicting nature, i.e., some criteria are more appealing than others for different decision makers, and thus, the selection process of the best alternative is vastly dependent on the decision makers’ preferences. This introduces discrepancy in the AoA process, which results from systematic errors introduced by the decision makers. This is common in a typical group decision-making scenario where many individuals are involved in the decision process, and thus, a method to aggregate the different evaluation viewpoints is mandatory. The ordered weighted averaging (OWA) operator is a mapping function that is used to aggregate different viewpoints. This thesis describes a new method to calculate the weight vector of the OWA operator based on the Laplace distribution. The proposed OWA operator is a new method for AoA to minimize discrepancy in alternative assessment, e.g., disagreement on the weight vector that leads to higher scores for the appealing criteria and smaller scores for the less interesting ones. The proposed OWA operator assigns smaller weights to both the higher and smaller scores, and thus, reduces the discrepancy in the AoA process. To prove the usefulness of the proposed operator, the calculated score is utilized in machine learning models as an explanatory variable for regression and classification problems and the results are compared to other OWA operators. The proposed OWA operator outperforms other operators in a breast cancer classification problem with an accuracy of 99.71%. Furthermore, a new model based on the calculated score and the Z-score is proposed for alternatives performance evaluation. The results of this method are illustrated using a case study for used cars performance ranking and evaluation with sensitivity analysis.
- ItemOpen AccessClinical Decision Support System with Adaptive Software Framework for Chronic Lymphocytic Leukaemia Cell Classification(2013-09-16) Mohammed, Emad; Far, Behrouz; Naugler, ChristopherThis thesis presents a new clinical decision support system (CDSS), which operates within an adaptive software framework and a tailored wrapper design pattern for chronic lymphocytic leukaemia (CLL) cell classification. The system goes through a sequence of steps while working with the lymphocyte images: it segments the lymphocyte with average segmentation accuracy of (97% ±0.5 for lymphocyte nucleus and 92.08% ±9.24 for lymphocyte cytoplasm); it extracts features; it selects from those features the relevant ones; and, it then classifies the selected features. The proposed system composite classifier model has a trust factor of 84.16%, accuracy of 87.0%, 84.95% true positive rate, and 10.96% false positive rate. The framework along with the wrapper pattern became a generic interface for any new algorithm. The framework built on top of the data-centric architecture which provides a great flexibility to the system design. The wrapper verifies the new algorithm interface against built-in test procedures.
- ItemOpen AccessDetecting Eye Diseases and Intraocular Lesions from Fundus Images Using Deep Learning Approaches(2023-12-20) Shakeri Hoosein Abad, Esmaeil; Far, Behrouz; Crump, Trafford; Mohammed, Emad; Kim, KangsooIn this study, the focus begins with addressing the critical issue of diabetic retinopathy (DR) detection, a leading cause of blindness globally, by using a combination of SHapley Additive exPlanations (SHAP) analysis and transfer learning ResNet50 model. Achieving impressive accuracy rates of 97% for binary and 81% for multi-class DR classification, the study demonstrates the potential of SHAP analysis to enhance interpretability and contextual understanding of prediction outcomes. Shifting the study to uveal melanoma (UM), an intraocular cancer with significant risks, the research used similar methodologies to predict UM, achieving a high binary classification accuracy of 82.5% in InceptionV3 model. The application of SHAP analysis once again highlights its value in shedding light on prediction rationales and improving result comprehension. The study further extends into the use of four distinct convolutional neural network (CNN)-based architectures for UM detection, emphasizing the manual collection and preprocessing of 854 RGB fundus images. Through transfer learning, DenseNet169 appears as the most accurate model, achieving 89% accuracy in binary classification of choroidal nevus (CN). Essentially, SHAP analysis continues to play an essential role in enhancing interpretability, offering detailed insights into the significant image regions influencing CN predictions. In conclusion, this study emphasises the power of combining deep transfer learning CNN-based models, and SHAP analysis to not only achieve robust predictive performance but also to address the critical challenge of interpretability in deep learning models, contributing significantly to the fields of medical image analysis and diagnostic decision-making.
- ItemOpen AccessA Practical Deep Learning Approach to Detect Aggressive Driving Behaviour(2022-01) Talebloo, Farid; Far, Behrouz; Mohammed, Emad; Sanati Nezhad, Amir; Moshirpour, MohammadAccidents while driving might result in minor injuries. Alternatively, it might result in a loss of life, which is highly detrimental to society. The loss of an expert due to fatalities can have a tremendous influence on humanity's scientific growth. Three factors can lead to accidents on the road: 1) The human, 2) the road, and 3) the vehicle. We look at the first element in our analysis, accounting for 93 percent of all accident causes. We will not look at the psychological aspects of driving behaviour in this study. The first step is to classify the vehicle; Self-driving vehicles and regular automobiles, both of which may be used to evaluate driving, are the two types of vehicles that can be checked. Aggressive driving behaviours have been identified as one of the most critical subcategories of human factors that contribute to accidents. To prevent road accidents, constant monitoring of drivers' driving behaviour can modify the driver's driving behaviour or notify the driver of a potential hazard. As a result, it is vital to devise a method of detecting aggressive driving behaviour. Aggressive driving is every day among American drivers. According to AAA Foundation for Traffic Safety data from 2019, approximately 80% of drivers displayed severe anger, hostility, or road rage while driving at least once in the preceding 30 days. Aggressive driving has been a significant source of concern for many road users. There are numerous methods for detecting aggressive driving behaviour, including changes in vehicle speed, lane shifts, eye and hand movement analyses, and others. We conducted this study using deep machine learning approaches rather than classic time series analysis methods. We analyzed roughly sixty similar publications to learn the procedures employed in the prior studies. The CNN was utilized in most publications to determine how to drive. We used RNN algorithms to execute this experiment since the vehicle GPS data is a time series. We employed an external test technique during the experiment that was not used in earlier studies that dealt with the same data set. The provided model produced satisfactory results incorporated in the dissertation's conclusion.