Information Sciences

Permanent URI for this collection

http://etd.aau.edu.et/handle/123456789/27

Browse

Now showing 1 - 20 of 779

Classifying Insider Threat from Electronic Mail Communication
(Addis Ababa University, 2016-06-22) Firesenbet Adela; Wondwossen Mulugeta
In a current interwoven global world the means of communication has been diversified. Electronic mail is one of the popular, simple and user-friendly for communication. The implication of this means of communication is reflected in various corners of the day to day activities of the modern world. Currently, email communication is set as a standard procedure for office communication in many organizations. Having the good face of such a communication approach, on the contrary unwanted distracting messages could bring institutional instability and even collapse. The objective of this research work is to classify the level of being insider threat using email text classification techniques from the electronic communication. In order to meet the stated objective, data mining algorithms in Weka 7.8 software has been used to classify the email texts. The experiment was conducted using 9808 negative and positive dictionary words identified by psychologists for training. For testing individual email files are used. The Enron higher officials email text was investigated after extensive text preprocessing techniques. The text preprocessing technique includes removal of email header, signature, alphanumeric character etc.SMO Classifiers are employed to manage the experiment. Therefore, the text email analyzed was categorized into negative and positive word counts then the negative word count was further classified into five stages of threat levels. Among twenty eight higher officials investigated at Enron Company, 22 of the employees were found at the exploration stage, one on exploitation stage, two on execution stage and three of them classified under escape stage. The evaluation of the classifier is acceptable and suitable for threat classification. Moreover, a court which was designated to investigate wire fraud, conspiracy and false audit report, convicted 3 of the officials spend in prison from 1.5 – 24.3 years. These individuals were classified under the escape stage of this study. Eventually, the output of this study indicates the promising use of text classification technique to trace and classify insider threats from email communication. Hence, further study and standardization of such a work could bring better result in organizational security and institutional functioning.
Mobile Banking and Mobile Money Banking Fraud Detection Using Machine Learning on Banks in Ethiopia
(Addis Ababa University, 2024-02) Daniel Manaye; Workshet Lamenew
Fraud is a criminal act with significant societal impacts, particularly in the financial sector. As digitalization expands rapidly in Ethiopia, the financial industry has suffered substantial losses, with an estimated 1.8 billion birr lost to fraud over the past four years. This study aims to explore and evaluate the effectiveness of SVM machine learning technique for detecting fraud in mobile banking and mobile money services within the Ethiopian banking sector. This research employs a quantitative experimental approach to investigate fraud detection in mobile banking and mobile money services using machine-learning models, particularly Support Vector Machines (SVM). A comprehensive literature review reveals that while mobile banking and mobile money have become essential in East African countries, including Ethiopia, there is a significant gap in research addressing fraud detection in this context. As the digital financial landscape evolves, the threat of fraud is becoming increasingly severe, posing a substantial challenge to the region's financial stability. This study aims to bridge this research gap by exploring and evaluating the effectiveness of machine learning (ML) techniques for detecting fraud in mobile banking and mobile money services. Utilizing the CRISP-DM framework for data mining, the study apply SVM supervised ML techniques to transaction data from these platforms. To address the class imbalance inherent in the data, under sampling techniques were employed, with the dataset split into training (80%) and testing (20%) sets after the necessary data cleaned and preparation based on the framework selected has been carried out. In this study, the data taken for analysis is the transaction data of mobile banking and mobile money this is because the fraudulent activities on one of the channel may come to the other as the services are having many similarities in nature. The performance of a Support Vector Machine (SVM) model was assessed using metrics such as Precision, Recall, and Confusion Matrix. Initial findings indicate that the model struggles with the class imbalance, which affects its overall effectiveness but still identify 51% of the fraudulent transactions. Despite these challenges, this study provides valuable insights into the application of machine learning for fraud detection in mobile banking and mobile money services within East African region where such practices are still emerging. This research contributes to the limited body of knowledge on fraud detection in the rapidly expanding digital financial services sector in East Africa, offering a foundation for future studies and practical applications in the financial section in the region and beyond.
Assessing Readiness of E-Learning in Five Ethiopian Resource Center (Cluster Lead) Universities from Teachers and Students Perspective
(Addis Ababa University, 2024-10) Fanuel Zegeye; Temtim Assefa
As the importance of e-learning for higher educational institutions grows, assessing the readiness of these institutions is crucial. The general objective of this study is to evaluate the e-learning readiness of five Ethiopian Public Resource Center (Cluster Lead) Universities from the perspectives of both teachers and students. Utilizing a descriptive research design, the study gathered information on the level of e-learning readiness through random and purposive sampling to ensure a representative subset from various departments and levels of experience. A total of 350 combined online and printed questionnaires were distributed to teachers and students, with 328 returned, yielding a response rate of 93.71%. The findings indicated that instructors showed greater readiness than students in areas such as technology, individual, and organizational e-learning readiness aspects necessary for successful implementation. While students demonstrated some readiness in individual technological skills essential for effective e-learning, they were not prepared in other critical readiness factors. Several areas were identified as lacking readiness and requiring additional improvement. In conclusion, despite some encouraging readiness factors that support the implementation of e-learning, significant challenges remain. This thesis recommends that the management of the five Ethiopian Public Resource Center (Cluster Lead) Universities actively enhance the areas that are ready but need improvement, as well as address those that are not prepared and require more attention. Additionally, it is crucial for the government to prioritize e-learning readiness to address the challenges encountered in this domain.
Identification of Information Seeking Behaviour: in Case of Ethio Telecom Customers
(Addis Ababa University, 2021-02) Mulunesh Belew; Daniel Alemneh
The aim of this study is examine the information need and information-seeking behaviour of Ethio telecom customers based in Bahir Dar town. A structured questionnaire was used to collect pertinent data from sample respondents. A total of 385 sample respondents were recruited using a systematic random sampling technique. The mean and standard deviation andcross-tabulation were used to test the relationship between dependent (Types of information needed,Source of information and Challenges encountered when seeking for Information) and independent variables (demographic characterstics of respondents and customer relationship with the company). Data were collected through personal-administered questionnaire. Descriptive analysis results revealed that website and contact centre are preferred information channels, but self-care application, IVR and mass media are not preferred. Interms of information need ethio telecom customers in Bahir dar town need information about almost all kind of products except Domain Name System, FAX, VPN and CRBT service. Availability of information sources, Lack of awareness about source of information, Reliability of credibility of the information source, Ability of sources to meet information, Affordability of the information sources, Difficulty in accessing both print and online material, Lack of accessibility of sources, Lack of time, Inadequate current information material were found as Challenges encountered when seeking for Information.
Applying Data Mining Techniques for Customers Segmentation and Prediction: the Case of Dashen Bank
(Addis Ababa University, 2023-06) Etsegenet Gebregiorgies; Million Meshesha
The banking sector has changed significantly in how it does business, putting a greater emphasis on contemporary technology to stay competitive. The banking sector has begun to understand how critical it is to build a knowledge base and use it to the bank's advantage in the field of strategic planning. Finding clients who are more likely to be interested in a product or service is a crucial task. Data mining has been widely used for customer segmentation and identification in order to predict potential customers for a given product or service. This study uses a six-step hybrid Knowledge Discovery Process model with the goal of applying data mining for the purpose of customer segmentation and prediction. The necessary data was gathered from the Bank's CBS database, and then pre-processing operations like data transformation and cleansing were performed in order to produce high-quality data for use in data mining with WEKA software. The goal of this thesis is to create a model that can be used to categorize Dashen Bank customers based on their transactional data and forecast which customers will be profitable for the bank. Since there are no predefined classes that describe the customers of the bank, the researcher uses clustering techniques (such as Kmeans, Filtered cluster and Farthest First) that resulted in the appropriate number of clusters for customer segmentation. K-means clustering, which divides potential customers based on their monthly credit turnover, produces the best descriptive model. By labeling the unlabeled data set as a result of clustering, classification algorithms like J48 Decision Trees, K Nearest Neighbor (KNN), and Naive Bayes can be used to build a model that allows for customer prediction. Researchers divide the data into three distinct clusters based on the transactional amount range by using a clustering algorithm. The labels "SMALL," "MEDIUM," and "CORPORATE" are applied to these clusters. "The range of transaction is the main distinction between these clusters. Experimental result shows that, out of the three algorithms, J48 decision tree with 70/30 test mode have the highest performance accuracy of 92.08%, which is selected for customer prediction. After consulting with experts, the data mining analysis revealed intriguing and unexpected attributes and patterns. The findings indicate that customers who exhibit basic deposit behavior with an overdraft facility are classified as corporate transactional customers. On the other hand, customers who hold a USD account are also categorized as corporate customers, which results in higher profitability for the bank. The study is based on only transactional data and customers are segmented on their monthly activities. To get 360 customers view, further research needs to be done towards coming up with customer profiling and customer relationship management (CRM) system
Ethio Telecom Airtime Credit Risk Prediction Using Machine Learning
(Addis Ababa University, 2023-06-01) Selahadin Nurga; Million Meshesha
Prepaid mobile consumers can airtime credit to use telecom services even after their balance has run out and pay for it later users will find this service useful and operators also make more money from it but there’s also a chance that subscribers wont pay their credits credits dank this study’s main focus is on how machine learning techniques are applied to ethio telecoms airtime credit service customers to evaluate credit risk. Ethiopia’s top telecom company ethio telecom needs to manage credit risk well to maintain financial stability and customer satisfaction the company can identify customers who are more likely to default on their airtime credit by using accurate credit risk Ethiopia’s top telecom company ethio telecom needs to manage credit risk well to maintain financial stability and customer satisfaction the company can identity customers who are more likely to default on their airtime credit by using accurate credit risk prediction which enables proactive measures to lower risks and boost financial performance the historical customer data include in this study include customer profile data call detail data loan history data and usage data preprocessing techniques are used before model training to handle missing values encode categorical variables and reduce features ensuring the quality and consistency of the dataset. In order to predict the credit risk associated with airtime this study used supervised machine learning algorithms four different machine loaming algorithms including naïve byes classifiers logistic regression random forests and k-nearest neighbors were trained and tested using a dataset of 1.168.000 ethio telecom prepaid subscribers performance evaluation metrics like accuracy precision retail and FI- score are used to assess each models efficacy using class balancing strategies the models robustness and gencralizabilizbility are also validated According to experimental results the random forest algorithm has successfully predicted airtime credit risk with 99% accuracy in order to identify customers who are highly likely to default on their airtime credit ethio telecom is able to take preventative actions with the help of this developed model like adjusting credit limits The study’s findings can fill the gap on how credit risk is predicted in the telecommunications industry and demonstrate how machine learning can improve risk management strategies and financial performance the proposed method can be used as a foundation for the development of automated credit risk prediction systems for ethio telecom and comparable organizations resulting in enhanced decision-making processes and reduced financial losses
Assessment of TB Screening and Referral Linkage among HIV Patients Attending ART Clinic of Butajira Hospital, SNNPR, Ethiopia
(Addis Ababa University, 2010-06-01) Shemdin Jemal; Wakgari Deressa
Background: Tuberculosis (TB) screening recommended [or people living with human iminunodeficiency virus (PLWHA) to facilitate early diagnosis and safe initiation of antiretrc viral therapy (ART) and Ionized preventive therapy (lPT). The interaction between TB and r-II ! infection is complex. HIV :infect on weakens the immune system and increases the susceptibility to TB. I-TIV increases the likelihood of reactivate on, re- infection and progression of tattle TB infection to active disease. Objectives: The aim of this study was to assess TB screening and referral linkage among HIV patients attending ART clinic of Butajira hospital. Methods: PL WH ~ who were enrolled in the lime priced from 1998 - 200 I E.C studied. Both quantitative and qualitative data collection methods were used to conduct the study. A total of 384 patient's charts that fulfill the inclusion _ criteria were select d by systematic random sampling technique for cross sectional stud y. For qua1itative study, by purposive sampling, 10 healthcare providers and program coordinators were interviewed to complement to the quantitative study. By purposive bivariate and multivariate ate analysis were used to determine the factors associated with TB diagnosis. The qualitative data were analyzed thematically. Result: Among the screened PLWHA, 97 (2 5.3%) were on INH and 224 (58.3%) of patients Were co-trimoxazole prophylactic therapy (CPT). Out of evaluated PL WHA 300 (78.1 %) were screened for TB d There was no TB screening for VCT clients who were tested I HIV positive at V T clinic. TPT & CPT were not provided for all eligible PLWHA. The referral linkages or HIV patients to ART clinic with intra-facility and inter-facility were through referral lips; however most of patient's referral slips were not documented. Routine TS screening for all VCT clients who are tested HIV positive is recommended to be screened at VCT clinic. I PT & CPT wail be provided for all eligible PLWIIA according to FMOH TB/HIV Implementation on guidelines. TS/HIV technical cOlm.1ittce en-IT ) at Hosp ital, Zonal and Regional level should be re-organized and strengthen according to national TB/HIV guidelines. Periodic evaluate on of TS and HIV programmers should be strengthen at all level. Regular supportive supervision, bi-annually and annually at program management level is recommended. t least one time and 84 (21.9 %), were never screened for TB. Active TB'. I was diagnosed in 69 (23%) patients; of which 20 (29%), 33 (47.8%), and 16 (23.2%) of cases were smear positive pulmonary TB, pulmonary negate and extra-pulmonary TB, respectively. There were significant association between TB case finding versus cough, dispend, chronic fatigue, night sweat and fever; however the association of weight loss and hemoptysis were not significant. Concerning referral linkages, 277 (72. 1 %) patients were linked to ART clinic from In-patties it, Medical-outpatient, TB clinic, PMT-CT, General VCT, Other outpatient and PICT with intra-referral slips. 107 (27.9%) patients were referred from other health facilities to ART clinic. Conclusions and recommendations: chronic fatigue, cough, night sweat, and fever were the most frequently appeared symptom complexes of1 B among the screened patient's charts. There was no TB screening for VCT clients who were tested HIV positive at V T clinic. TPT & CPT were not provided for all eligible PLWHA. The referral linkages or HIV patients to ART clinic with intra-facility and inter-facility were through referral lips; however most of patient's referral slips were not documented. Routine TS screening for all VC'T clients who are tested HIV positive is recommended to be screened at VCT clinic. IPT & CPT wail be provided for all eligible PLWI-iA according to FMOH TB/HIV Implementation on guidelines. TS/HIV technical committee-IT ) at Hosp ital, Zonal and Regional level should be re-organize d and strengthen according to national TB/HIV guidelines. Periodic evaluation of TS and HIV programmers should be strengthen at all level. Regular supportive supervision, bi-annually and annually at program management level is recommended.
Factors Affecting the Practices of Clinical Health Workers in Manual Patient Record Registration Process
(Addis Ababa University, 2014-02-01) Shimelis Shiferaw; Ayele Belachew
Introduction:- The objective of the HMIS would be to record information on health events and check the quality of services at different levels of health care. One of the services is patient record registration process which could be affected by different factors. Objectives: - The general objective of this study was to assesses those factors like knowledge, attitudes, practices and demographic variables that might affect the practice of clinical health workers on manual patient record registration processes, in the case of Addis Ababa Public Hospitals. Methods:-An institution based cross sectional study was conducted using mixed (quantitative and qualitative) techniques. Two stage sampling methods were used to select the clinical health workers. Self administered questionnaires; interview and patient record review were used as data collection tools. A simple random sampling technique was used to access patient records. Percentile, frequencies, chi-square test and logistic regression were used for analysis. The study was conducted from February 2013 until December 2013. Results and Conclusion:- Regarding the practices 44.1% of the clinical health workers have poor practices on manual patient record registration process. Educational level, sex and knowledge level of the participants were found to be factors that affect manual patient record registration processes. From multivariate logistic regression the health officers show 10.3 time good practices than specialists. Diploma nurses have 3. 6 times good practices than the specialists. And also the B.Sc. nurses have shown 3.32 times good practices than the specialists. Regarding gender and practices on patient record registration process, males have 1.83 times good practice than females. To improve the practices of clinical health workers with high education level providing performance based motivation/ incentives may help in boosting the practices of patient record registration processes. Increasing number of nurses and implementing three shift systems might reduce the work load of the nurses. Providing sufficient attention and sufficient budget allocation by Addis Ababa Health Bureau is mandatory to improve data quality in hospitals.
English - Afaan Oromoo Machine Translation: an Experiment Using Statistical Approach:
(Addis Ababa University, 2009-04-01) Sisay Adugna; Andreas Eisele
Machine translation (MT) refers to the use of a machine for performing translation task which converts text In one natural Language Into another Natural Language It can have many applications like cross-linguistic Informant retrieval and speech to speech translator systems It can also assist professional translators by producing draft quality output Thai reduces cost that would be Incurred if translation and typing was done manually from scratch English Is the lingua franca of online Information and Afaan Oromo Is one of the most resource scarce languages For this reason, monoflngual Afaan Oromoo speakers need to use documents Witten other languages, among whelp English IS the most popular one To satisfy this need, translation of the English documents to Afaan Oromoo and thus making these on line documents available In Afaan Oromoo is vttal in addressing the language barrier thereby reducing the effect of digital divide. Therefore, this thesis Is focused on the development of a prototype English-Afaan Oromoo machine translation system using statistical approach. i.e. without explicit Formulation of linguistic rules.as this approach involves low cost and swiftest way Available these days Using limited corpus of about 20.000 bilingual sentences. a Transition accuracy of 17 74% was achieved
Amharic Part-of-Speech Tagging Using Hybrid Approach (Neural Network and Rule-Based)
(Addis Ababa University, 2008-09-01) Solomon Asres; Sebsibe Hailemariam
The accumulation of information in this electronic age is rapidly increasing. Yet we have very little intelligent tools that will help individuals manage this giant information. Natural Language Processing researches are looking closely at this problem and try to build systems that can understand natural languages. Part-of-speech tagging is one attempt in the effort of understanding human languages. It is the assignment of a category to a word which indicates the role of the word in a given context. There are a lot of part-of-speech taggers for many languages but is not for Amharic language. This study proposes a hybrid method of Neural Network and Rule-based approach for tagging Amharic words. So this method is based firstly on Neural Network and then anomaly is corrected by Rule-based approach. Back Propagation algorithm and Transformation- Based learning method are adopted for the development of Amharic tagger. Building the tagger with hybrid approach can improve the performance of the tagger. This study sets better Amharic tag sets, large size corpus and uses two methods for better accuracy. To evaluate the proposed method, a number of experiments have been conducted. A large number of data are used to train and test the tagger. The experimental result of this thesis work indicates that 91 % and 94% for rule-based and neural network tagger, respectively. But the result reaches to 98% when the experiment has been conducted on the hybrid tagger
Mining Vital Statistics Data: the Case of Butajira Rural Health Program
(Addis Ababa University, 2011-06-01) Tadesse Beyene; Million Meshesha
Data milling is a relatively new field whose major objective is to ex tract knowledge hidden in large amounts of data. Vital statistics data offer a fertile ground for data mining by providing valuable source of information regarding the health status of a population. one of the most important public health functions is monitoring of a pulsation’s health Swills. At all levels of the health deli very structure a well organized health information system is vital for identifying the health needs of populations and for planning. implementation and monitoring of health interventions. The aim of this study is to discover knowledge that can be used to gain insight in to various aspects of mortality in the selected rural area of the country. The study explores the death aspect of the vital statistics data in the Blaire Rural health Program- BRHP database at butajira, Ethiopia. A data mining tool called weak is used build predictive model of 95,220 cases over an eighteen-year r period. A historical cohort study analyst is of vital statistic is conducted. It follows a IDM process modeling. This study apply classification algorithm , such as to extract interesting knowledge from temporal data on BRI-I!> database. The results obtained in the study contain valuable new information. These results com-eyed some interesting findings. The class frication algorithm reveals that the res lust indicates for the BRHP-' dataset, over 90% accurate results are possible for developing class frication r les that call be used in prediction From this result the researcher concludes that the vital statistics data can help to predict using the application of data mining classification technique given the limitation of this study. III general. the result from this study is encouraging
Assessment of Use of Information Technology in Pharmacies and Drug Stores in Addis Ababa, Ethiopia
(Addis Ababa University, 2010-06-01) Tadesse Gebre; Mulugeta Betre
Back ground: Advance In information technology (IT) provides easy use and access to exploit its benefits. Information technology is currently important in improving healthcare delivery system. They improve efficiency, effectiveness, and reduce medication errors . However, the use and access of this in formation technology at pharmacies and drug stores level among d rug dispensing professionals is little known in Ethiopia, particularly in Addis Ababa. Objective: To assess the u se and access of information technology among pharmacists and druggists in pharmacies and drug stores in Addis Ababa and to identify the factors that affecttheusc of this technology. Method: A cross sectional survey was conducted in pharmacies and drug stores in Addis Ababa. The quantitative data were collected using pretested and self Administered questionnaire. The study Vas complimented with in-depth interview. Data were entered and analyzed by SPSS version 15.0. Result: A total of 257 pharmacist and druggists pa reticulated the study. The current means of giving service to customers .in pharmacies and drug stores was about t 93.0% paper -based (manual ) system and 7.0% were u sing computer system. Only 30(1 2.0%) professional had in tenet access in pharmacy/drug s tore. The most preferred source to obtain drug information was combination of printed sources and d rug inserts (manual system) 169(68.1%). Use of IT in pharmacy/drug store was poor. Conclusions and recommendations: The study indicated poor utilization status of IT for pharmacy practice service. The findings indicated the need for creating awareness among professionals In giving more skill oriented and also forma l in service IT re la ted trainings for the professionals. Further, the drug profess Iona t raining centers as well as other stakeholders should cons id reapproving the IT facilities for drug dispensing professional is to achieve better universal access and see of IT so as to improve healthcare delivery system, particularly pharmacy practice
Assessment of Information Security Culture in Public Hospitals in Southern Nations Nationalities and Peoples Region/SNNPR/: The Case of Hadassah Referral Hospital
(Addis Ababa University, 2010-06-01) Temesgen Gebrasilase; Lemma Lessa
Background: - Traditionally, most of the researchers and expels in the field of intonation security believed a technological solution to address majority of intonation security issues. However, contemporary studies have shown that non-technical solutions including the human behavior and processes are as important as technical solutions in safeguarding organization intonation assets. Objective: - To assess the intonation security culture of Hadassah Referral Hospital in order to improve the existing information security culture of the hospital. Methods: - A cross-sectional survey was conducted in Hadassah Referral Hospital from March-April 2010. A total of314 study subjects were emancipated for questionnaire and in-depth interview. The data was collected by using a non-structured per-tested questionnaire, in-depth interview and document analysis. The quantitative data was analyzed by using SPSS version 15.0 and the qualitative data was analyzed manually. Results: - It was found that, 66.9% of the study participants have no knowledge about intonation security. About 63% of the respondents reported that management does not assist to the implementation and incorporation of intonation security in the hospital. The result of the study also showed the absence of well written and documented intonation security policy in the hospital. Conclusion and recommendations: - The study showed that majority of the respondents in the hospital has no knowledge about intonation security. The study also revealed that there is a lack of commitment from top management for the incorporation and implementation of intonation security in the hospital. Hence, it is recommended that there is a need to provide extra-education and training on intonation security issues for health care providers, administrative staffs and medical students. It is also recommended that the hospital management need to increase their support and commit enough resources to bring acceptable level of intonation security culture and practices in the hospital.
Grammar Checker for Amharic Language Using Probabilistic and Rule-Based Approach
(Addis Ababa University, 2019-07-01) Tsedeniya Kinfe; Martha Yifiru
Grammar checker is an natural language processing (NLP) application which validates a sentence grammatically based on a predefined rule. Grammar checkers have been developed using different techniques for different languages. Investigation on the development of Amharic grammar checker was conducted by [1] using two approaches ( i.e. rule based and statistical approach) independently to handle simple and complex Amharic sentence respectively. The rule-based approach performs better compared to the statistical method in detecting grammatical error for simple sentences. Purpose of this research is, therefore, to investigate the application of a probabilistic method with rule-based approach in the development of Amharic grammar checker. In addition, in the previous study bi-gram and trigram LM were used for handling complex sentences. Tri-gram LM achieved better performance but it fails to detect long-distance disagreement within a sentence. Therefore, this paper investigated a long distance agreement by building higher order n-gram LM i.e. 4-gram and 5-gram LM. Moreover, the use of dependency parsing (DP) in Amharic grammar checking has also been investigated. To conduct the experiment, POS tagged data was used from [2] and this corpus is used for investigating the probabilistic method with rule-based approach and higher order n-gram LM. We have developed automatic POS tagger and chunkier to easily identify the subject, object, and verb of a sentence. SRILM toolkit is used to develop tri-, Quadra-, and pent-gram LM. The Treebank corpus from [3] is used to investigate dependency parser. To validate and optimize the Treebank Malt optimizer is used. To train and parse new sentences Malt Parser toolkit is used. Each parsed sentence is grammatically analyzed based on the crafted rules. To evaluate the parsed sentence LAS, LA, and UAS metrics are used and are found in Malt Val toolkit. The result achieved using probabilistic method with rule-based approach is 72% of accuracy while the higher order n-gram method resulted in 65%, 67% and 65.6% for tri-, Quadra-, pent-gram language models, respectively. Whereas dependency parser scores 81.5%, 94.2% and 84.4% in LAS, UAS, and LA respectively. The overall accuracy achieved by DP is 84.7% in detecting grammatical errors.
Formulating an Information Security Policy Framework for Ethiopian Banking Industry
(Addis Ababa University, 2021-06-01) Yoseph Getu; Temtim Assefa
Today's organizations rely heavily on information and information technology (IT) for the mere function of business operation s and stand out from the competition, especially for an organization like banks, as they acquire sensitive data. So these data must be protected at all times against any type or form of attack. Organizations fall victim to such attacks from poorly crafted, redundant, and weak information security policies (ISPs). This study aims to answer the research question; what are the core values needed develop an information security policy for the Ethiopian banking industry? Furthermore, this ca n help to determine what security issues exist and the weaknesses and Vulnerabilities of the organization. The study explored international information security governance frameworks and best practices; and chooses ISO audit checklist, combined with the researcher's experience to develop the framework. The researcher employed a qualitative research approach. Both primary data; through interviews and secondary data; document analysis are collected and used. A thematic analysis method is used in this research for analyzing the data. To analyze the data QDA MINER liter v2 .0.8 tool is used. Twenty four (24) core elements (codes) under ten (10) master themes; management of security, Acceptable use, data classification level, physical/environmental security, intellectual property right, protection of malicious software, continuity of operations, contracts of employment and services, information asset management, and Access control are identified. The study findings show that the core elements availability in the Surveyed banks vary. Furthermore, they are at different position in handling the security of their systems. An entry-level ISP framework is formulated and evaluated . The Framework will be the basis of the organizations IS program and serve as a guideline for Creating an ISP.
A Case Based Reasoning Knowledge Based System for Type II Diabtes Management: Case of Desire Referral Hospital
(Addis Ababa University, 2011-11-01) Zewditu Sisay; Getachew Jemaneh
Diabetes mellitus (DM) is a common chronic disease around the world in which the boas not produce insulin (type I diabetes) or does not properly use insulin (type II diabetes ) the study investigated the potential of case-based reasoning (CBR)approach for type II DM treatment CBR is an approach to artificial intelligence that is intended to mimic an approach that people typically use to solve problems this is the use of past experiences to reason about new situations In order to acquire the knowledge the researcher conducted unstructured interviews with domain experts selected through expert sampling and other relevant documents then the knowledge is modeled using tree like structure called ladders patient history cards from dessie referral hospital in outpatient department (OPD)were the primary sources of cases. case attribute identification and weight assignment were done with the help of domain experts the case-based contained 42 type II DM cases and stored in plain text file (attribute value pair vector) the prototype is built using Jcolibri a software artifice that promotes software reuse for building CBR systems JCOLIBRI employed nearest neighbor retrieval algorithm for retrieval and propose the most similar cases for reuse manual revision is by the domain expert in order to adapt a stored case’s solution for a new case there is always incremental learning through retaining newly solved cases The prototype performance is evaluated through statistical analysis and user feedbacks recall and precision were the main statistical performance measures using leave-one-out cross validation testing proportion the retrieval performance of the prototype showed average value of 69% recall and 46% precision domain experts were also evaluating the prototype using certain criteria The main objective the research is to design and build CBR knowledge based system that retrieves relevant previously stored cases and proposes appropriate solution the developed prototype scores promising performance and user acceptance
Conflict Analytics: a Predictive Model to Forecast Violent Conflicts in Ethiopia for Improved Early Warning Systems
(Addis Ababa University, 2023-06) Meheret Takele; Wondwossen Mulugeta
In the context of an increasing number of violent political conflicts across nations and societies, understanding and predicting violent conflicts that lead to significant economic, social and humanitarian consequences is both an academic interest and a moral obligation. Research on conflict prediction is critical to offer input for policymakers to see potential conflicts and devise strategies for conflict early warning. However, the application of data analytics tools to analyze the dynamics of violent conflicts in Ethiopia has hardly existed. With the motivation to fill this knowledge gap, this research aims to develop a model using supervised machine learning algorithm that can best forecast violent political conflicts in Ethiopia in terms of the dominant conflict types/categories and regions at risk of conflict incidents. Methodologically, this research has employed an experimental research design and adopted the CRISP- ML framework. Predictive analytics tools, as well as three algorithms (random forest, gradient boosting and Gaussian naive Bayes), are used. Open-source software called Jupyter Notebook is used for analysis. The research combined past and recent conflict incidents data with political, economic, social and environmental data from 2007 to 2022. After collecting data on both independent and dependent variables from open databases of research institutes and international organizations, models were built, compared and assessed using a new dataset of variable values projected for the next five years (2023- 2028). The finding shows that the Gradient Boosting machine learning algorithm has a better performance than Random Forest and Gaussian naïve Bayes in predicting the location and types/categories of conflict classes individually. In predicting types of violent conflicts, while Gradient Boosting has a testing accuracy of 57%, both the Random Forest and Gaussian naïve Bayes has 55%. Yet, in terms of predicting the location of conflict incidents, all have 48% testing accuracy. However, when both type/category and location of conflict incidents were considered together the predictive performance of the selected algorithms declined to 30% testing accuracy. In conclusion, the performance of conflict-predicting models is determined by whether the classes (type and locations) are merged and predicted concurrently or separately and independently. The best performing algorithm(57%), such as the Gradient Boosting machine learning algorithm, predicts that Ethiopia will continue to be at risk of violent conflicts for the coming five years, 2023- 2028. The location and type of violent political conflicts shift yearly. With its original findings, this research can make valuable contributions to policy-making and academic fields, such as Information Science, Conflict Studies and others.
Improving Healthcare System Though Process Mining at Tikur Anbessa Specialized Hospital
(Addis Ababa University, 2024-10) Mesele Awulachew; Million Meshesha
The quality of hospital services depends on the effective execution of healthcare processes, which encompass a variety of clinical and non-clinical activities performed by diverse resources. These processes are dynamic, complex, and multi-disciplinary, necessitating a deep understanding of improvement. Process mining offers promising techniques for visualizing and analyzing healthcare processes to enhance efficiency and quality. This study therefore aims to apply process mining techniques to discover healthcare process models, identify deviations and inefficiencies, and optimize resource allocation within the hematology department of a healthcare organization. The process model followed in this research includes data extraction from Tikur Anbessa Specialized Hospital (TASH), data preprocessing using aggregation, temporal approach and simple heuristic, process discovery using heuristics mining and inductive mining, and model evaluation based on fitness, precision, generalization, and simplicity. Control flow, performance and organizational analyses are also conducted, followed by validation of findings through expert collaboration. The analysis highlights the most common pathway for hematology patients begins with a laboratory request, followed by a laboratory test, then a hematology diagnosis, and finally a prescription, highlighting the interconnectedness of these processes. However, discrepancies between the number of laboratory requests and completed tests, coupled with an average test duration of 32 days, significantly above World Health Organization (WHO) benchmarks, reveal inefficiencies, particularly in resource allocation. Comparative analysis using heuristic and inductive miners demonstrated the inductive miner showing superior fitness, precision and simplicity, with the heuristic miner achieves a slightly higher in generalization. Social Network Analysis (SNA) identified strong interdepartmental interactions, especially in the diagnosis and radiology departments. The proposed process improvement framework was well-received, achieving an overall mean evaluation score of 4.2 and a Cronbach’s alpha of 0.747, indicating its reliability. These findings emphasize the complexity of healthcare processes and the importance of continuous improvement through integrated systems. Future research should address challenges in data quality issue to further enhance the utility of process mining in healthcare settings.
Detection of Pneumonia Using Deep Learning Approach from X-Ray Images
(Addis Ababa University, 2024-09-27) Melat Alemayehu; Million Meshesha
Pneumonia classification is developing an automated system capable of distinguishing between three distinct categories—normal, non-pneumonia lung diseases, and pneumonia—using chest X-ray images. Pneumonia is a severe lung infection that can lead to significant morbidity and mortality if not promptly diagnosed and treated. However, the challenge arises from the fact that other lung conditions, like chronic obstructive pulmonary disease (COPD) or tuberculosis, can present with symptoms and X-ray findings similar to pneumonia, making accurate differentiation crucial. The objective of this study is to design and evaluate a machine learning model that can accurately classify these categories, thereby aiding in the early and accurate diagnosis of pneumonia versus other conditions. This classification task is vital for improving clinical outcomes, reducing diagnostic errors, and ensuring that patients receive appropriate treatment as quickly as possible. The primary objective of this research is to develop an effective model for pneumonia classification using deep learning approach. By categorizing chest X-ray images into three classes—non-pneumonia (other lung diseases), normal (healthy lungs with no signs of disease), and pneumonia—we aim to enhance diagnostic accuracy and improve radiologist performance. To achieve this, we utilize a data set collected from Tikur Anbessa Specialized Hospital, comprising 3000 chest X-ray images, with 1000 images per class to ensure balanced representation before augmentation; and also, we used a ratio of 80:10:10 (80% training, 10% validation, and 10% testing) splitting ratio. We employed an experimental research approach, selecting three state-of-the-art pretrained models—InceptionV3, ResNet50, and VGG16—for transfer learning, alongside constructing a custom CNN-L5 model. Through a series of experiments, we investigated various image prepossessing techniques, including re-sizing, normalization, and image enhancement, to optimize model performance. We explored different combinations of epochs, batch size, and learning rates for all models, while also experimenting with fine-tuning the pretrained models. The most effective model for pneumonia classification was then selected based on these trials. The experimental results showed that the CNN-L5 model, trained on enhanced image data sets with 30 epochs, a batch size of 64, and the inclusion of dropout, achieved superior performance with a classification accuracy of 96.8%. This highlights the importance of using appropriate preprocessing techniques and optimizing model architecture to achieve high performance in pneumonia classification. However, this study was limited to data from a single hospital, which may not represent the diversity of patient populations, imaging techniques, and conditions found in other healthcare settings. Consequently, the model's performance may not generalize well across different environments. To address this limitation, future work should incorporate data from multiple hospitals to improve the model's robustness and broader applicability.
Data Quality Improvement Model for Awash Insurance’s Information System
(Addis Ababa University, 2024-09) Naol Gemeda; Melkamu Beyene
The study analyzes and improves data quality issues in the information system of Awash Insurance S.C. through providing a comprehensive model known as Insurance Data Quality Improvement Process Model (IDQIPM). The approach contributes to operational efficiency and decision-making reliability of the company (AIC) by addressing data quality problems such as instance duplication, data value inconsistency, inaccuracy, and incompleteness issues of its Information system data. Design Science Research Methodology with qualitative approach is employed in this study. A data quality assurance model that can enhance data quality within Insurance’s underwriting operation, one of the core business processes of the company and that faces most of data quality problems is proposed. The models effectiveness has been attempted to be shown in demonstration section of this paper which brought about encouraging results in terms of enhancing data quality within Awash Insurance S.C.'s insurance information system. However, future research might improve the model so as to enhance various data quality dimensions which is not addressed in this study.

Browse

Recent Submissions