http://ir.haramaya.edu.et//hru/handle/123456789/46 Tue, 02 Jun 2026 09:04:26 GMT 2026-06-02T09:04:26Z http://ir.haramaya.edu.et//hru/handle/123456789/8465 Automatic Question Generation from Afaan Oromo Text UsingDeepLearning Jemal Abdela; Million Meshesha (PhD); Jemal Abate (MSc) In the modern digital era, the availability of electronic content in multiple languages has grownconsiderably. Nevertheless, generating questions from these materials remains a labor-intensiveand time-consuming process, particularly for low-resource languages like Afaan Oromo. Whilesignificant progress has been made in Automatic Question Generation (AQG) for languages suchas English, Chinese, Amharic and Somali. there is a notable lack of technology with NLPonAfaan Oromo QG because Afaan Oromo has the unique linguistic characteristics. This studybridges this gap by developing an AQG system for Afaan Oromo using deep learning models, specifically LSTM, Bi-LSTM, and GRU. To build the model a dataset of 5,000 paragraph-question-answer triples was meticulouslycurated from Afaan Oromo textbooks and educational resources. The dataset underwent preprocessing steps such as tokenization, normalization, and word embedding using Word2Vec. A deep learning model with an attention mechanism was employed to generate questions bydesign science research methodology, Among the models evaluated, the Bi-LSTMdemonstratedthe highest performance, achieving a training accuracy of 95.3% and a validation accuracyof 92.5%. The LSTM model also performed well, with a training accuracy of 92.31%and a validation accuracy of 91.02%, while the GRU model showed marked improvement after hyperparameter tuning, reaching 88.0% training accuracy and 87.0% validation accuracy.The results indicate that the Bi-LSTM model is the most effective for generating both factoidandnon-factoid questions in Afaan Oromo. Future research should explore transfer learning frompre-trained models, expand the dataset through collaborations with educational institutions, andintegrate advanced neural architectures like Transformers to further enhance performance andquestion quality. This study makes a significant contribution to the field of Natural LanguageProcessing (NLP) by pioneering AQG for Afaan Oromo, use as input for QA system, providingafoundation for future research and practical applications in education and language preservation. 138p. Sat, 01 Mar 2025 00:00:00 GMT http://ir.haramaya.edu.et//hru/handle/123456789/8465 2025-03-01T00:00:00Z http://ir.haramaya.edu.et//hru/handle/123456789/8464 Automatic Question Generation from Afaan Oromo Text UsingDeepLearning Jemal Abdela; Million Meshesha (PhD); Jemal Abate (MSc) In the modern digital era, the availability of electronic content in multiple languages has grownconsiderably. Nevertheless, generating questions from these materials remains a labor-intensiveand time-consuming process, particularly for low-resource languages like Afaan Oromo. Whilesignificant progress has been made in Automatic Question Generation (AQG) for languages suchas English, Chinese, Amharic and Somali. there is a notable lack of technology with NLPonAfaan Oromo QG because Afaan Oromo has the unique linguistic characteristics. This studybridges this gap by developing an AQG system for Afaan Oromo using deep learning models, specifically LSTM, Bi-LSTM, and GRU. To build the model a dataset of 5,000 paragraph-question-answer triples was meticulouslycurated from Afaan Oromo textbooks and educational resources. The dataset underwent preprocessing steps such as tokenization, normalization, and word embedding using Word2Vec. A deep learning model with an attention mechanism was employed to generate questions bydesign science research methodology, Among the models evaluated, the Bi-LSTMdemonstratedthe highest performance, achieving a training accuracy of 95.3% and a validation accuracyof 92.5%. The LSTM model also performed well, with a training accuracy of 92.31%and a validation accuracy of 91.02%, while the GRU model showed marked improvement after hyperparameter tuning, reaching 88.0% training accuracy and 87.0% validation accuracy The results indicate that the Bi-LSTM model is the most effective for generating both factoidandnon-factoid questions in Afaan Oromo. Future research should explore transfer learning frompre-trained models, expand the dataset through collaborations with educational institutions, andintegrate advanced neural architectures like Transformers to further enhance performance andquestion quality. This study makes a significant contribution to the field of Natural LanguageProcessing (NLP) by pioneering AQG for Afaan Oromo, use as input for QA system, providingafoundation for future research and practical applications in education and language preservation. 138p. Sat, 01 Mar 2025 00:00:00 GMT http://ir.haramaya.edu.et//hru/handle/123456789/8464 2025-03-01T00:00:00Z http://ir.haramaya.edu.et//hru/handle/123456789/8129 CONSTRUCTING A PREDICTIVE MODEL FOR CREDIT RISK ASSESSMENT IN THE BANKING SECTOR USING MACHINE LEARNING TECHNIQUES GELETA BEGNA; Million Meshesha (PhD) A commercial bank is a type of financial intermediary that takes deposits and uses them to fund lending operations. Credit risk occurs when a debtor fails to make a loan or other credit-related payments. Banks often regulate and manage risk through the discipline known as risk management. Creating and implementing a technology that can effectively support risk management is crucial. This study used Awash Bank Share Company as a case study to examine how machine-learning techniques support loan risk assessment. Following the acquisition of credit data from Awash Bank, appropriate features were selected and data preparation procedures, including data preprocessing, were accomplished. Recursive Feature Elimination (RFE) was utilized for attribute selection during data preprocessing, and efforts were made to clean the data by handling missing values and correcting errors. The important attributes that were identified are Loan Type, Amount Granted, Mode of Repayment, Principal, Interest Rate, Days Overdue, Net worth, Current ratio, Quick Ratio, Debt to Asset ratio, Debt to Equity Ratio, Net profit margin, Liabilities, Total Capital, Total Asset, and target class Status. To address data imbalance and control model overfitting, SMOTE and L2 regularization were applied. This study employs two data splitting techniques (Percentage split and 10-fold cross-validation). For constructing a prediction model machine learning algorithms, such as MLP (Multi-Layer Perceptron), SVM (Support Vector Machine), and LR (Logistic Regression) are employed to identify and classify credit Status into Pass or Loss categories. According to the experimental results, the MLP classifier achieves a higher classification accuracy of 99.35% because MLP's ability to learn complicated patterns as well as its independence from data sequence dependencies led to its excellent performance in this context and hence we suggest MLP for constructing an applicable system for credit risk assessment. So, this study successfully constructs a predictive model with higher accuracy for Credit Risk Assessment in the Banking Sector to identify customers' credit repayment ability by using Machine Learning Techniques. However, the study constructs a predictive model based on the data obtained from one bank, Awash Bank. It is therefore our recommendation to collect data from different banks so as to construct a generic model. 93 Sat, 01 Jun 2024 00:00:00 GMT http://ir.haramaya.edu.et//hru/handle/123456789/8129 2024-06-01T00:00:00Z http://ir.haramaya.edu.et//hru/handle/123456789/8115 DEVELOPING ISLAMIC LEGAL KNOWLEDGE-BASED SYSTEM USING RULE BASED REASONING Abubeker Ali Adem; Dr.Million M. (Ph.D.); Tilahun S. (Assistant Professor) The objective of this research primarily focused to capture tacit and explicit knowledge from domain experts and existed documents in order to support the decision making processe of the domain experts and to provide a solution for existing problems, such as problem related to a societies social (i.e. racism, inappropriet marriage and divorce), economical (support the process of financing), and political (i.e. islamic penal laws for crime and punishment) issues. Based on extracted knowledges rules were formulated to develop knowledge-based system for islamic legal system inorder to address the difficulty of getting informations and knowledges for the problem stated above. After developing the knowledge based system, the system were demonstrated and evaluated to test the performance and user acceptance of the proposed knowledge-based system. The study followed Design Science Research Methodology. And also, in order to evaluate the proposed knowledge based system questionary were used. The knowledge’s collected from experts and codified sources were represented in the form of rules and facts. The inference engine is used for reasoning by comparing the rules and facts stored in rule base with user specific problem or facts which are stored in working memory, so the inference engine provides solution for problem related to marriage, divorce, commercial dealing and islamic banking service, racism, islamic penal laws, and islamic rules and regulations in order to advice, recommend, direct, and support decision making of an experts and users. And also the result obtained suggested that the proposed knowledge-based system can play an important role to the process of building family, maintain peace and fight crime and racism, and support economy of country. One of the major limitation was unable to follow a proper approaches such as questionary along with interview before developing the knowledge-based system in order to identify and understand existing problem and applicability of knowledge-based system. Furthermore, the proposed rule-based knowledge-based system is limited on reasoning and unable to learn from its experience so if it intgrete with case based reasoning more complex problems can be handled 127 Thu, 01 Aug 2024 00:00:00 GMT http://ir.haramaya.edu.et//hru/handle/123456789/8115 2024-08-01T00:00:00Z