http://ir.bdu.edu.et/handle/123456789/10145 Sat, 13 Jan 2001 07:32:54 GMT 2001-01-13T07:32:54Z http://ir.bdu.edu.et/handle/123456789/16480 AMHARIC SPEECH TO ETHIOPIAN SIGN LANGUAGE TRANSLATION SELAMAWIT, BELAY Ethiopian Sign Language (EthSL) is the primary means of communication for the deaf community in Ethiopia. The dearth of accessible tools that convert Amharic speech into EthSL, however, makes it difficult for hearing and deaf people to communicate. The goal of this study is to create a deep learning model for translating Amharic speech to video sign language. To close this gap, this study attempts to create a deep learning-based Amharic speech-to-video sign language translation model. Four deep learning models were compared: Long Short-Term Memory (LSTM) networks, Variational Autoencoders (VAEs), Recurrent Neural Networks (RNNs), and Convolutional Neural Networks (CNNs). Via hyperparameter optimization, LSTM achieved the highest accuracy, increasing from 91% to 97%. In order to ensure model performance, interpretability, and usability, the study used a design science methodology. To increase the model's decisionmaking process's transparency, integrated gradients were used. A Flutter-built mobile application including the trained LSTM model was evaluated for usability using surveys. Results show that as compared to previous deep learning models, the LSTM-based model considerably increases the accuracy of Amharic speech-to-EthSL translation. The prototype improves accessibility for the deaf community by enabling real-time EthSL video creation from Amharic speech. By offering a dependable and understandable solution, encouraging inclusivity, and setting the stage for future developments in Amharic speech-to-sign language translation, this research advances assistive technology. For this thesis, four deep learning algorithms were investigated: Long Short-Term Memory (LSTM) networks, Variational Autoencoders (VAEs), Recurrent Neural Networks (RNNs), and Convolutional Neural Networks (CNNs). After hyperparameter optimization, LSTM outperformed the others, increasing its accuracy from 91% to 97%. Using a design science methodology, this study guarantees interpretability and usability in addition to model performance. Transparency in Amharic speech-to-video sign language translation was improved by using Integrated Gradients to explain the model's decision Sun, 01 Jun 2025 00:00:00 GMT http://ir.bdu.edu.et/handle/123456789/16480 2025-06-01T00:00:00Z http://ir.bdu.edu.et/handle/123456789/16479 Virtual Machine Recommendation for Resource Utilization in the Cloud and Edge Kidus, Fiker Mon, 04 Nov 2024 00:00:00 GMT http://ir.bdu.edu.et/handle/123456789/16479 2024-11-04T00:00:00Z http://ir.bdu.edu.et/handle/123456789/16478 ESTIMATING SOFTWARE MAINTENANCE COST USING MACHINE LEARNING ALGORITHMS JEMMAL, AREGA Software maintenance costs can represent up to 67% of total expenses in the Software Development Life Cycle (SDLC), often surpassing 50% across all phases. This highlights the importance of accurate cost estimation for effective project planning and resource management. Traditional approaches, such as expert judgment and algorithmic models, often fail to address the complexities arising from changing requirements, aging codebases, and evolving technologies. This research aims to create a machine learning-based model for more reliable software maintenance cost estimation, utilizing data gathered from domestic software companies through interviews with project managers and analysis of archival project records. Factors affecting costs were identified through a literature review and refined via interviews. Various machine learning algorithms, including Linear Regression, Ridge Regression, Decision Trees, Random Forest, Support Vector Regression (SVR), Gradient Boosting Regressor, and XGBoost, were assessed, with XGBoost emerging as the most effective. It recorded the lowest error rates (MAE: 437.51, RMSE: 620.05) and the highest R² (0.95). To enhance transparency and build stakeholder trust, SHAP and LIME were used to explain the model's predictions. The research concludes with the development and integration of the XGBoost model for practical software maintenance cost prediction, thereby improving accuracy in cost estimation and resource management for software projects. Keywords: Software Maintenance Cost Estimation, Machine Learning, XGBoost, SHAP and LIME Fri, 01 Nov 2024 00:00:00 GMT http://ir.bdu.edu.et/handle/123456789/16478 2024-11-01T00:00:00Z http://ir.bdu.edu.et/handle/123456789/16476 Enhancing Pediatric Healthcare Using the Integration of Software Framework and Machine Learning ESHETIE, GASHAW YIGIZAW In pediatrics, infectious disease is a subspecialty that addresses the diagnosis, prevention, and treatment of infections in children aged from birth to 21 years of age. Given their developing immune systems and frequent exposure, children are especially vulnerable to infections. Globally, infectious diseases have a significant impact, causing millions of deaths each year. The advancement of emerging technologies, such as machine learning, has gained new momentum to fight against pediatric infectious diseases. This study investigates the application of machine learning (ML) in enhancing the diagnosis and treatment of pediatric infectious diseases, aiming to improve healthcare outcomes for the pediatric population. We employed a quantitative research design approach, combining an experimental phase to develop and fine-tune ML models like KNN, NB, SVM, LR, RF, and XGBoost with a survey method to evaluate the effectiveness of the machine learning integrated software framework. We emphasize meticulous data preprocessing, utilizing the K-Nearest Neighbors (KNN) imputation method for handling missing data and the Synthetic Minority Oversampling Technique (SMOTE) for addressing data imbalances. These preprocessing steps are critical for enhancing model performance and accuracy in complex medical applications. Additionally, z-score normalization is applied to standardize datasets, ensuring stable and reliable ML model outcomes. After conducting the experiments, we found that Random Forest performed best and integrated it into a framework designed for practical use in pediatric healthcare settings. This work integrates SHAP (SHapley Additive exPlanations) into a random forest model to enhance transparency and build trust among healthcare stakeholders. A software framework incorporating these explainable models was developed to improve both usability, understandability, and transparency. We performed a usability test with clinicians, resulting in a SUS score of 74.25, which corresponds to C and acceptable on the grade and acceptability scale, respectively. Using a random forest model, we achieved 0.97 of accuracy in predicting pediatric infectious diseases, employing a 90/10 traintest split, 5-fold cross-validation, and grid search hyper parameter optimization technique. With this result integrated with the framework, our study contributes to facilitating the analysis of patient data and identification of healthcare trends, which supports the clinicians, limited in number and unable to perform a good deal of diagnoses in a short period of time. Keywords: pediatric infectious diseases, machine learning, explainability, software framework, system usability scale. Tue, 01 Oct 2024 00:00:00 GMT http://ir.bdu.edu.et/handle/123456789/16476 2024-10-01T00:00:00Z