Browsing by Author "Mulugeta, Wondwossen (PhD)"

Now showing 1 - 12 of 12
  • No Thumbnail Available
    Item
    Amharic Character Recognition System for Printed Real-Life Documents
    (Addis Ababa University, 2010-06) Teshager, Abay; Mulugeta, Wondwossen (PhD)
    Optical Character Recognition (OCR) is an area of research and development where a system is made to recognize characters from printed documents. Cultural considerations and enormous flood of printed documents motivated the development of OCR across the world. Unlike other scripts, OCR development for Amharic Characters has been started in 1997 at SISA (School of Information Studies for Africa). Some developments have been made in recognizing various types of machine-printed, typewritten and handwritten Amharic documents. However, Amharic character recognition is still an area that requires the contribution of many research works. There is a need to enhance its performance on real-life documents such as the ‘Addis Zemen’ Amharic newspaper, the Bible, the ‘Federal Negarit Gazeta’ and the fiction ‘Fiker Eskemekabir’, which have a number of artifacts (mode of writing, condition of the input page, printing process, quality of paper, presence of extraneous markings, resolution and quality of scanning etc.) that affect the performance of the recognizer. One such area, OCR technology has been investigated more for real-life Amharic degraded documents. For the recognition to be successful, robust techniques in detecting and removing various noise types are investigated and validated. During experimentation of the applicability of algorithms and approaches for the problem at hand, MATLAB Image processing Toolbox and neural network classifier on MATLAB Neural Network Toolbox is used. The wiener adaptive filtering method for noise removal, Otsu global thresholdingmethod for binarizing the digitized image, linear interpolation techniques for normalization and hit-and-miss morphological analysis for thinning are found to work very well for the problem of interest. In due course, the performance of the line segmenter is found to be 100%. The rate of segmentation for basic and labialized characters turns out to be 98.28% and 100% respectively for training character sets, 98.55% and 100% respectively for testing character sets. For classifying the features generated, an artificial neural network approach is implemented. The neural network is trained with eight samples taken from real-life documents. The performance of the developed system is tested with documents taken from real-life documents. Accordingly, an average recognition rate of 96.87% for the test sets from the training sets and 11.40% recognition rate is observed for the new test sets. The segmentation algorithm used in the current study worked reasonably for basic and labialized characters. But it fails to segment special character |v|, punctuations and numbers. In general, observation of the test results show that the performance of the system is greatly affected by the similarity of the shape of Amharic characters and effect of the application of noise removal for cleaning highly degraded document images. Such challenges require to further explore an invariant to shape feature extraction techniques and advanced noise detection and removal algorithms. Based on the results, further research areas are also recommended.
  • No Thumbnail Available
    Item
    Big Data Governance Framework for Ethio Telecom
    (Addis Ababa University, 2022-01-14) Shewangizaw, Bereket; Mulugeta, Wondwossen (PhD)
    With the digitization of most operations, the popularization of various social media channels, blogs, deployment of various types of devices, sensors, hand-held smart devices, wearables, and the explosion in Internet usage, large volumes of data are being generated on a regular basis. Today, businesses are searching for ways to successfully manage and optimize these large datasets due to a massive amount of information being swapped every day and the growing need to make better data-driven decisions. Big data refers to huge data sets with large, more varied, heterogeneous, and complex structures that are expensive to retrieve, examine, and visualize using conventional data processing technologies. One of the most pressing issues in dealing with big data is the adoption of suitable big data governance frameworks to customize big data in a sufficient manner to enable quality data access for effective knowledge extraction using machine learning techniques. It also aspires to outline the blueprint that governs the storage and processing of data from owners and consumers in a truthful manner within the applicable regulatory landscape. Ethio telecom is one of the major big data custodians in Ethiopia that lacks such insight. The main objective of this work, therefore, is to explore and propose a big data governance framework for Ethio telecom based on governance activities related to data handling in Ethio telecom networks. The proposed framework consists of three domains and within the domains, there are nineteen components identified to formulate the structure. Towards that end, the research adopted Design Science Research as a research approach coupled with a mixed methodology for data collection and analysis. Towards achieving the objective, a review of literature on big data, big data analytics, big data management, big data governance, and previously designed big data governance frameworks were explored with the aim to develop a suitable framework and identify common characteristics and shortcomings in the available big data governance frameworks. Primary data were obtained through survey questionnaires and key informant interviews to maintain organizational context and domain-specific big data governance specifications. As governed by the chosen research strategy, these two contributions (qualitative and quantitative data) are integrated to design a detailed big data governance framework suitable for the Ethio telecom setting. Expert validation was then used to assess the proposed big data governance framework. As a result, the research process and study results are thought to be acceptable, indicating the usability and applicability of the proposed framework.
  • No Thumbnail Available
    Item
    A Comparative Study of Automatic Language Identification of Ethio-Semitic Languages
    (Addis Ababa University, 2018-06-06) Bekele, Rediat; Mulugeta, Wondwossen (PhD)
    The dominant languages under the family of Ethio-Semitic languages are Amharic, Geez, Guragigna and Tigrigna. From the findings of the language identification studies on European languages, there is a conclusion that most classifiers performance reached the accuracy of 100%. Local and global studied confirmed that Naïve Bayes Classifier (NBC) classifier does not reached the accuracy level of 100% in language identification especially on shorter test strings. Comparative Language Identification studies in European languages shows that Cumulative Frequency Addition (CFA) performs close to 100% accuracies better than the NBC classifier. The purpose of our study is to assess the performance of CFA as compared to NBC on Ethio-Semitic languages, to validate the research findings of CFA and NBC classifiers, and recommend the classifier, language model, evaluation context and the optimal values of N that performs better in language identification. In this research we have employed and experimental study to measure the performance CFA and NBC classifiers. We have developed a training and test corpus from online bibles written in Amharic, Geez, Guragigna and Tigrigna to generate 5 different character based n-gram language models. We have measured the classifiers performance using under two different evaluation contexts using 10-fold cross validation. F-score is used as an optimal measure of performance for comparing classifiers performances. The classifiers commonly exhibited higher performance when the length of the test phrase grows from a single word to 2, 3 and beyond to reach an F-score measure beyond 99%. Both classifiers performed similarly under each context corresponding to the language models and n-grams tested. The language model, fixed length character n-grams with location features, exhibited highest performance in F-score for both classifiers under each evaluation contexts on test strings as short as one word length. N=5 on Fixed length character n-grams with location features language model is the optimal value of N whereas N=2 is the optimal value for the remaining language models on both CFA and NBC classifiers and evaluation contexts. Based on our findings CFA is a classifier that performs better as compared to NBC as it is founded in sound theoretical assumptions and its performance in language identification.
  • No Thumbnail Available
    Item
    Design Considerations Framework for Mobile Based Service to Promote Reproductive Health Information for the Youth-at-Risk in Addis Ababa
    (Addis Ababa University, 2021-09-12) Hagos, Yonas; Mulugeta, Wondwossen (PhD)
    It is common for teenagers to face many puberty-related challenges as they transition to adulthood, and these challenges are worsened by a lack of access to reproductive health information. Youth are the most affected in current times because of various health disturbances. For youths to make better choices, they must be informed about critical life issues, including reproductive health information. Mobile technology is now a vital part of everyday life and is increasingly popular among younger generations, opening new possibilities for providing relevant mobile-based services for obtaining information. Mobile platforms have recently made health information more accessible and shared. The research project aims to improve the development of reproductive health information services for youth aged 15-24 in Addis Ababa through mobile-based services with an emphasis on design considerations. Most of the solutions and mobile-based services based on mobile technology are copied from elsewhere, with little or no consideration for the communities and individuals who will use these services. Throughout the process of creating the model and the prototype for mobile-based reproductive health information service, youth were engaged in different activities as part of the design process as design partners. In a systematic manner, design science methods were used as the methodology. The design science research phases guided the design process of the service prototyped here, from ideation to co-design. The different methods youth-at-risk use to access information, especially reproductive health information, were identified. Technology and existing practices for accessing reproductive health information were also identified. According to the results, the youth use mobile phones for sharing and receiving health-related information. In the current day and age, youth have access to reproductive health information and services, but they must travel long distances to reach some services by foot or by vehicle. It is recommended that when developing any community-based solution, all stakeholders need to be included in the co-design process to ensure an ideal solution for users that is useful and relevant. By involving the user in the design process, innovations become more easily adopted by the user and technology is easy to adopt. User contributions are needed since they contribute knowledge, experiences, and ways of doing things that improve the solution. The study recommends that services be developed in the future based on the findings herein.
  • No Thumbnail Available
    Item
    Designing Knowledge Management Framework for Facilitating Indigenous Craft Knowledge Preservation and Transfer the Case of Cloth Production at Shiromeda
    (Addis Ababa University, 2021-09-04) Mulualem, Chalachew; Mulugeta, Wondwossen (PhD)
    This study aimed to investigate the Design knowledge management framework that preserves, and transfers indigenous craft knowledge of Cloth production practices in local communities of Shiromeda, Addis Ababa, Ethiopia. The paper talks around the significance of indigenous Knowledge, particularly the traditional cloth production indigenous knowledge, of the Ethiopian cultural cloth production sector. The study distinguishes the sources where the indigenous knowledge resides and examines the current knowledge sharing culture between different partners in connection to traditional cloth production; then adjusts different concepts from pre outlined partnered Knowledge Management (KM) models to plan an Indigenous Knowledge Management Framework (IKMF) custom-made to Ethiopian cultural cloth setting. Mixed approach is utilized to examine existing gaps that will likely be started by different relevant attributes. Qualitative approach is employed for interviewed data analysis and quantitative approach for analyse and drive conclusion from questionnaire data. This examination substantiated by different sources to assemble knowledge, examine and at last created an IKMF that envelops a cyclic Indigenous Knowledge preserve and transfer processes with the thought of different factors extracted from collected data. Accordingly, 30 key respondents were chosen through purposive sampling and were used for the quantitative data collection. For the qualitative data semi structured interviews were used for data collection where 26 respondents were interviewed and analysed using thematic content analysis moreover qualitative data were analysed using thematic approach. The results showed that the local communities shared IK of traditional cloth production by using traditional practices. Poor knowledge sharing and preservation culture, lack of trust, political dimensions or social status, poor recognition of IK holders, lack of an awareness are the barriers of effective management of IK of traditional cloth production practices in the study community. Hence, we develop an ICK framework that will support indigenous knowledge preservation and transfer. The framework also facilitates keeping the traditional cloth production knowhow and activities among community, craftsman and stakeholder organizations so as to facilitate preservation and transfer to new generations.
  • No Thumbnail Available
    Item
    Designing web based Social Health Insurance Information System for Ethiopian Health Insurance Agency
    (Addis Ababa University, 2016-06) Alemu, Natnael; Mulugeta, Wondwossen (PhD)
    Social health insurance scheme is a mechanism of distributing or pooling of risk among individuals. The scheme can be used to reduce individual burdens and to provide service at a subsidized cost among payroll based formal sector. The Social health insurance scheme involves three stakeholders these are the insurer, the insured and the service provider. The use of information and communication technology to create cooperative working environment between the stakeholders in the scheme can improve the accessibility and use of health care services at a point of service irrespective of geographical and time location. This can help to improve the use and access of citizens to the primary and advanced health care services. The application of ICT to the major business process of health insurance enables, an ease of access for the various health insurance services the Ethiopian health insurance agency provides to the employees. The main objective of the project is to assess the existing paper based social health insurance system and design a user friendly web based social health insurance information system for Ethiopian health insurance agency The project was limited geographically in Addis Ababa and conceptually to the designing of the web based social health insurance information system in the formal sector. The project employed an object oriented system analysis and design technique with different data collection tools i.e. (interview, observation, relevant document review) to collect the data required for system design. Analysis and design of the proposed system was done using the UML, Microsoft Visio 2013. While MySQL Server and Dreamweaver 5.5 development environments have been used to develop the database and the prototype of the web based system respectively. The designed and developed information system of Social health insurance for the Ethiopian health insurance Agency comprises of different functionalities like employee(the insured) registration, validation, claim adjudication, insurance registration components which will help to transform the paper based manual system to efficient, increased geographical coverage and time unconstrained system. The web based social health insurance information system could enhance accessibility of services and information or data transaction possible with the reduction of the unnecessary time consuming process. Most importantly the users access to the system be user friendly, save data, have appropriate notifications and manage incomplete data display with appropriate layout and access path.
  • No Thumbnail Available
    Item
    A Framework Development for Application of Data Warehouse in the Ethiopian Banking Industry the Case of Dashen Bank
    (Addis Ababa University, 2022-01-29) Tesfaye, Zelalem; Mulugeta, Wondwossen (PhD)
    Data warehouse is the fastest growing technology integrated with Business intelligence to make a better decision using analytics on the current as well as historical data of the organization. The purpose of this research is to develop a framework for the application of Data warehousing technology in the Ethiopian banking industry, the case of Dashen bank. The research followed the Kimball dimensional data modeling technique to show the conceptual and logical data model. For this particular Applied research, a qualitative research methods are believed to be more suitable in describing the process of developing the data warehouse and Data collection is mainly based on Focus Group discussion between the staff members of the IT department, Observation and document assessment to evaluate the current status of data integration and readiness of the bank for the application data warehouse. With this the readiness of the bank is evaluated, the data concentration area in the bank’s current system is also identified, source databases and the type of data with these systems is also identified and these findings analyzed and represented using the UML class diagram on the conceptual level. Based on the conceptual level, conformed dimensional tables and Fact tables with their attributes and measures are identified. The logical design section of the study shows the relationship of facts and dimensions using star schema of Kimball dimensional modeling. The finding of this research tried to shows the benefits of the application of data warehousing technology in the Ethiopian banking industry using the conceptual and logical design of the data warehouse in the case study organization.
  • No Thumbnail Available
    Item
    A Hybrid Cloud Computing and Service Environment for Ethiopian Banks
    (Addis Ababa University, 2017-06-05) Solomon, Kebede; Mulugeta, Wondwossen (PhD)
    Information technology is the common element for all the industries this days. Consequently any important change in this area will have a direct or indirect impact on small and large scale organizations. One of the technologies that have an influence on IT is cloud computing. Due to the importance and sensitive nature of data/applications used by financial institutions, and other factors such as, competitions, changing customer, and line of business needs, Ethiopian banks face problems in providing necessary services to customers, allies, and employees, by using the ideal channel at any time. The main goal of this research is to examine how Cloud computing could change the way services are provided to customers, employees job satisfaction for Ethiopian banks. In order to do so, a proposed cloud model has been introduced based on existing cloud models and services in combination with the opinion of IT experts from selected Ethiopian banks. Literature review and Interview were used as the research methodology for this thesis. Detailed study regarding cloud computing, services, models, and its security has been done. Moreover interview was selected for gathering opinions from the selected IT leaders and experts of the banks. The general architecture, its features that form cloud infrastructure, best practices in other countries, were also explored in this paper. In this research I proposed Hybrid Cloud Computing model consisting Private Cloud and Community Cloud model. This proposed model would have the advantage of effective, efficient, reduced IT investment cost, reduce time to maintain system failures, and improve employee’s satisfaction for delivering better banking services to customers.
  • No Thumbnail Available
    Item
    Information Security Management Framework for Effective Implementation of Integrated Financial Management Information System (IFMIS) the Case of MoF
    (Addis Ababa University, 2020-08-06) Mekonnen, Yafet; Mulugeta, Wondwossen (PhD)
    The main purpose of the study is to propose Information Security management framework for integrated financial management information system (IFMIS). In this study, MoF was selected using purposive sampling that issues service for different financial sectors around in Addis. The target population constituted 108 employees, the IFMIS and IT staffs located at MoF were included to be part of this study. Data was collected by means of questionnaire; interview and group discussions and analyzed using descriptive statistics. The analyses include frequency distributions, tables, figures and Narrative description. 108 questionnaires were distributed and 84 (78.8%) were returned. In addition to the questionnaires, observation and document review was made to strengthen the respondents‟ view. Accordingly, the data is processed using IBM SPSS V 20.0 Statistical tool. The framework that is proposed extracted from ISO security standard, NIST cyber security framework, literatures, and supported by findings from survey conducted in the MoF. The components are interwoven and all together support implementation of effective security solutions. The study shows that the financial information security management framework and practice is not well maintained to address the MoF information security managements with associated to the IFMIS system. In general, the study shows that there is no standard to security, technical challenge management associated with the financial sectors. The study recommends that the management should involve on any aspect of the IFMIS project to improve the efficiency and minimize risks and technical challenges, the Ministry should have standard information security management framework and risk management techniques and policy to minimize and manage the risk and system. One of the best ways to make sure employees will not make costly errors in regard to information security is to institute organization-wide security awareness initiatives that include, but not limited to face-to-face and multi-media based awareness, techniques that can be fairly inexpensive to implement such as posters, do and don‟t lists and warning banners. These methods can help ensure employees have a solid understanding of the organization security policy, procedure and best practices. With the intention of elaborating on the underlying research that produced it, the proposed ISM framework for IFMIS was presented and discussed in detail – all the components, sub components, as well as the processes followed in preparing the framework. Finally, recommendations are given for the Ministry to act in short and long-term basis to improve the information security management awareness of its employees and in turn improve better information security management practice in the IFMIS.
  • No Thumbnail Available
    Item
    Predict the Major Factors that Helps to Predict Employee Turnover in Government Organization Using Machine Learning:- the Case of Ethiopian Federal Court
    (Addis Ababa University, 2020-05-12) Atinaf, Eristie; Mulugeta, Wondwossen (PhD)
    Nowadays, Employee turnover is a serious issue in organizations. It affects the time, productivity, and stability of the given organizations. Employees are very important that helps the organization get success and gain revenue. So, Organizations need to know the key issues that the reason for employee turnover. Prediction models are highly associated with human resource management to identify the employee turnover patterns from employee previously recorded data. The objective of this research is to design a model and predicting staff turnover using a machine learning approach in the Ethiopian Federal court organization. For prediction three classification models namely, random forest, logistic regression and gradient boosting tree were used. The total datasets from the three federal court organizations were 3610 both active and terminated.For evaluate the prediction classification models the researcher was use confusion matrix, recall, precision and roc-curve to measure the performance of the classifiers. After evaluation, from the three classification models the finding shows that the best classification model is gradient boosting tree with an accuracy of 87.5%. Additionally, from the study it is found that the factors responsible for employee turnover are:-experience, salary, age and employee’s number of year service are the most significant factors. The factors martial and gender were low predictor variables on employee turnover in the federal court organization.The study concludes that the most reliable and accurate classification model to predict employee turnover isan ensemble – based learning technique gradient boosting tree that was found as the most suitable classifier for building the predictive model.
  • No Thumbnail Available
    Item
    Students’ Placement Prediction Model: A Data Mining Approach
    (Addis Ababa University, 2017-06-02) Getachew, Meseret; Mulugeta, Wondwossen (PhD)
    The main objective of Higher Education institutions is to provide quality education to students. To achieve highest level of quality of education institutions may apply discovery of knowledge for prediction placement of the students’, consider resource capability and performance of the students. Thus researcher initiated to undertake study on students’ placement into different available departments using data mining technique and to propose a predictive model. The study attempted to build a predictive model for student placement prediction and identify interesting rules from the generated model by applying data mining techniques. The study been carried out using hybrid data mining methodology processing. The targeted data set used for the study was the students’ placement and high school score of students’, who joined Addis Ababa University during 2015/16 and 2016/17 Academics years. The data is acquired from AAU registrar office and NEAEA. The original dataset consist 34 attributes and very 11320 instances. Thus, to make the initial data appropriate and manageable for the data mining exercise, data preparation task was undertaken. Decision tree and rule induction classification techniques using J48, REPTree and PART algorithms were applied for the experimentation. The experiments were conducted using six scenarios for each algorithm and the outputs of the experiments were used for comparison of the models based on the set evaluation criteria. After the test design was defined and the dataset separated into training and test dataset, the model was built the training set and its quality was estimated on the separate test set. As a result, the test run using PERTree algorithm has registered best accuracy which is 82.045%. The generated rules were interpreted, analyzed and the discovered knowledge was evaluated against the existing knowledge base and domain expert’s validation. Based on the findings of this research work, we can conclude that improved students’ placement in to various departments can be done using data driven predictive model with acceptable.
  • No Thumbnail Available
    Item
    The Use of Data Mining to Predict the Loan Repayment Risk: the Case of Oromia Credit and Saving Share Company
    (Addis Ababa University, 2018-09-04) Feyissa, Ketema; Mulugeta, Wondwossen (PhD)
    Data Mining is the process of extracting useful patterns from the huge amount of database and many data mining techniques are used for mining these patterns. Data mining is still a technology of having great expectations to enable the organizations to take more benefit of their huge data bases. Recently, one of the remarkable facts in microfinance institute is the rapid growth of data and this microfinance data is expanding quickly without any advantage to the organization for decision making. The main aim of this research work is utilizing of data mining by developing classification model in order to predict customer loan repayment behavior for loan risk that could help for better decision making and maximize the benefit of the microfinance from organizational datasets. In this research the applicability of classification data mining techniques to implement customer loan repayment prediction model in Oromia credit and saving share company have been explored within the approach of CRISP-DM process model. After understanding business objectives of the organization, customer profile data are extracted, collected, cleaned, transformed, integrated and finally prepared for experimentation with the classification algorithm to develop a prediction model. The final dataset prepared for experimentation have 147,285 customer profile instances. The findings of this study revealed all the models built from J48 Decision Tree classifier, Naïve Bayes classifier and Neural Network have high classification accuracy and are generally comparable in predicting customer loan repayment. However, comparison that is based on their performance accuracy suggests that the J48 model performs slightly better in predicting customer loan repayment with classification accuracy of 98.89%. In this study the following attributes: customer follow up, purpose of loan, distance of customers from microfinance center and amount of loan disburse are the most interesting attributes in determining customer loan repayment prediction. The result of this study used efficiently to model and predict customer loan repayment. Based on the findings of the study, we recommend that microfinance institutions should adopt data mining to enhance their performance. The organizations need to make sure that there is enough data to analyze as well as assure quality of data. Organizations should ensure that the analysts are trained well and deduct the correct information which serves the purposes of the problem in the first place.