Mekelle University Institutional Repository
Discover scholarly works, research outputs, and institutional publications.
The Mekelle University Institutional Repository is a digital collection of scholarly and research outputs created by the university's faculty, students, and researchers. This repository provides open access to a wide range of materials, including articles, theses, dissertations, conference papers, books, and more."
By making our research available, we aim to
- Increase the visibility and impact of research conducted at Mekelle University.
- Promote knowledge sharing and collaboration within the academic community.
- Preserve and disseminate valuable scholarly works for future generations
Communities in DSpace
Select a community to browse its collections.
- A central archive for Mekelle University’s institutional abstract books from academic and research conferences.
Recent Submissions
test
(2026-02-02) test
HATE SPEECH DETECTION FOR TIGRIGNA LANGUAGE ON SOCIAL MEDIA USING DEEP LEARNING
(Mekelle University, 2024-10-28) Letebrhan Abay Gebrekiros
Hate speech on social media poses a significant challenge to online safety and social harmony, with far-reaching implications for individuals and communities. Social media platforms are frequently misused for harassment and targeting, with hate being expressed in various forms, such as sexism, racism, and political intolerance. Tigrigna, a widely spoken language in northern Ethiopia and the official language of Eritrea, holds significant cultural and social importance. Despite the growing number of Tigrigna speakers active on social media, research on detecting hate speech in Tigrigna remains scarce. This thesis investigates the design and development of a hate speech detection system for Tigrigna, leveraging a deep learning approach using Bidirectional Long Short-Term Memory (BiLSTM) and Long Short-Term Memory (LSTM) networks. The system was trained on a dataset consisting of 8,583 annotated social media posts, with 2,610 labeled as hate speech and 5,973 labeled as hate-free speech. The performance of the proposed model on the unseen test data gives an accuracy of 92% using the LSTM and 91.5% using Bi-LSTM models. These results suggest that further optimization can be done to enhance the effectiveness of the proposed method on the same dataset. The thesis concludes with recommendations for future research and practical applications.
Malaria Infection Detection Using Image Processing and Deep Learning Method
(Mekelle University, 2025-11-28) ABDIRAHMAN ALI AHMED
Malaria remains a critical global health challenge, particularly in developing regions, contributing to substantial morbidity and mortality worldwide. Timely and accurate diagnosis is essential for effective treatment and disease control. This thesis presents a novel approach to malaria detection using advanced image processing and deep learning methodologies. Specifically, convolutional neural networks (CNNs) were applied to the analysis of microscopic blood smear images, enabling precise identification of malaria parasites. To overcome the limitation of small datasets, extensive image augmentation techniques were employed to expand the dataset artificially, increasing its diversity and enhancing the CNN model’s generalization and performance. The augmented dataset was used to train the CNN models, which were evaluated through various performance metrics, including accuracy, recall, precision, and F1-score. Additionally, the confusion matrix, receiver operating characteristic (ROC) curve, precision-recall curve, and learning curves were utilized to demonstrate the efficacy of the proposed method. The CNN-based deep learning architecture was fine-tuned with parameters such as 100 epochs and a batch size of 128 for 64x64 image inputs. The model achieved an accuracy of approximately 97.57%, precision is 97.12%, recall is 95.84%, and the F1-Score is approximately 96.48%. and an AUC-ROC score of 99%, indicating exceptional capability in distinguishing between parasitized and uninfected samples.
This research underscores the potential of integrating image processing with deep learning for automated malaria diagnosis, offering a robust and efficient detection system suited for resource-constrained environments. By contributing to global malaria control efforts, this work advances the field of medical diagnostics and paves the way for future applications of deep learning in detecting other infectious diseases.
IMPACT OF DISTRIBUTED GENERATION ON PROTECTION DEVICE IN THE CONTEXT OF DISTRIBUTION NETWORK
(Mekelle University, 2025-12-15) Selemon Mesfin Gebreyowhans
The increasing integration of distributed generation (DG) into low voltage distribution networks has introduced significant protection performance challenges. Traditional non directional over current relays, designed for unidirectional power flow, often malfunction under DG operation due to changes in faults current magnitude, direction, and source contribution. This research investigates the impact of DG on relay performance and proposes effective mitigation strategy using coordinated and graded non directional relays. A 15kv radial distribution network with a total capacity of 24 MVA was modeled and simulated in DIgSILENTPowerFactory. A 6MW inverter based DG unit (solar PV and wind hybrid) was integrated at bus 47 to analyze its impact under various fault types (L-G, L-L, L-L-L) and locations. Three simulation scenarios were evaluated.1) based case (without DG), 2) DG integrated case, 3) mitigated case (with DG and 11 coordinated non directional relays).
The performance was assessed using key metrics: fault current, relay operating time, coordination time interval (CTI), selectivity, directional behavior, and unwanted/missed trips. Results indicated that DG integration changed fault current, reversed current direction, and led to protection blinding, sympathetic tripping, islanding risk and loss of selectivity. After applying the mitigation scheme with 11 graded relays, selectivity fully recovered, and no false trips and missed trips occurred. The mitigation thus demonstrated that DG induced issues can be eliminated using only non-direction relays, provided proper coordination and setting adjustments are made. Furthermore, a practical method for current transformer (CT) ratio selection was developed to ensure adequate sensitivity without saturation under DG fault conditions.
The key contribution of this research lies in showing that reliable DG integration can be achieved through analytical coordination and relay grading, avoiding the need for expensive directional or adaptive relays. The findings are highly relevant for distribution utilities in developing regions seeking low cost and technically viable protection solutions for networks with moderate DG penetration (<=25%).
Assessing the impact of urban-rural linkage in terms of local construction material flow: the case of Mekelle and surrounding areas
(Mekelle University, 2026-02-16) Haftay Tsegay Nere
Rapid urban expansion in Mekelle is driving the extraction of manufactured sand and gravel from surrounding rural areas, leading to increased bareland. This study examined the impacts of urban–rural linkages through construction material flows between Mekelle and nearby rural communities. A mixed quantitative and qualitative research design was applied. Used both primary and secondary data collected through different methods, including questionnaires. Random and purposive sampling methods involved 80 participants. The study was conducted at seven crusher sites in Hareko and Messebo tabias, purposively selected based on post-conflict functionality, proximity to Mekelle, and rural administrative locations. Also, three sand and gravel trading centers in Mekelle were selected based on their reliance on materials sourced from the study areas. The analysis adopted a cradle-to-gate system boundary within the broader cradle-to-grave framework due to data limitations. Results showed that annual production reached 86,680 m³ of manufactured sand and gravel, of which 15% generated as byproduct. The material flow chain supported 159 rural and urban residents through wage labor and trading activities. However, socio-environmental impacts were identified, including health risks, carbon emissions (2.23 kg C02 /t), and soil degradation. Although mitigation measures were agreed upon, weak regulatory enforcement prevented their full implementation, except for water spraying, which reduced dust emissions by 70% annually (492.41kg/year) but led to raised concerns over unsustainable water extraction. The study highlights policy gaps and recommends stronger regulatory enforcement to ensure sustainable resource management, alongside future research on why environmental policy in Ethiopia has weak in implementation.