Mekelle Institute of Technology

Permanent URI for this communityhttps://repository.mu.edu.et/handle/123456789/53

Browse

Filters

20253

Settings

Has files: YesSubject: search.filters.subject.Decision Tree

Search Results

Now showing 1 - 3 of 3
  • Item
    DEVELOPING PREDICTIVE DATA MINING MODEL FOR GRADE 11TH STUDENT PERFORMANCE: IN-CASE OF ELU WOREDA SECONDARY SCHOOL
    (Mekelle University, 2025-08-24) BEDADA DEMTEW
    Education is a key factor for achieving continuing economic progress. Many students are getting less in their result for many reasons. More academic institutes now store massive student educational and related data. A large number of students will enter every year. So, the demanding growth of data in education sectors continues. Handling and analyzing such a high amount of raw data manually for performance evaluation creates dissatisfaction, boring and unsuccessful. Manually handling and evaluating such a big amount of data for performance evaluation leads to discontent, boredom, and failure. To put it another way, traditional methodologies are overly complex and difficult to analyze and evaluate. To solve such kind’s issues, we use a data mining technique, which is a set of machine learning algorithms, to examine the data .In other words automated discovery of previously unknown, valid, novel, useful, and understandable patterns in school databases. The study's major goal is to use data mining to create a predictive model for student academic performance in Oromia National Region State's South West Shoa zone Elu woreda Teji Senior Secondary School and Asgori secondary schools. This can greatly was support policymakers, planners, and education providers working on the control of student performance. The methodology used for this research will a hybrid six-step CRISP Knowledge Discovery Process. The essential data would be gathered from a school data ware house created specifically for student result purposes, would be collected from 2011 to 2016 E.C. The researcher used three popular data mining algorithms (J48 Decision Trees, Random forest and Naive Bayes Classifier) to develop the predictive model using a larger dataset (4000 cases). The researcher used 10-fold cross-validation and percentage split test mode for data mining algorithms of the three predictive models for performance comparison purposes. The results indicated that the J48 Decision tree with10-foldcross-validation mode and Random Forest with 85% split testis the better predictor due to the nature of data which is categorical for which this two algorithm is better. Both algorithm (the J48 Decision tree with10-foldcross-validation mode and Random Forest with 85% split test) has 100% accuracy on the given school dataset and Naïve Bayes came out to be the second with an accuracy of 97.5%
  • Item
    Diagnosis of Diabetes Using Data Mining Techniques
    (Mekelle University, 2025-09-24) AMANA TESHI GEMMEDA
    Diabetes is a chronic metabolic disorder characterized by elevated blood glucose levels due to the body's inability to produce or effectively use insulin. It is a significant and growing health concern worldwide, affecting millions of individuals and posing a considerable burden on healthcare systems. Diabetes can lead to various complications, including cardiovascular disease, kidney damage, nerve damage, and vision impairment. Early and accurate diagnosis of diabetes is crucial for effective management and prevention of complications. Traditional diagnostic methods rely on clinical symptoms, medical history, and laboratory tests. However, with the advancements in technology and the availability of large healthcare datasets, data mining techniques have emerged as a promising approach for diabetes diagnosis. Diabetes diagnosis plays a crucial role in effective disease management and prevention of complications. This study explores the application of data mining techniques for diabetes diagnosis, aiming to improve diagnostic accuracy and support healthcare decision-making. Four classification models (Decision Tree, SVM, Naïve Bayes, and Neural Network) are constructed and evaluated using a comprehensive diabetes dataset. Performance evaluation metrics, including TP rate, FP rate, accuracy, precision, recall, and F1-score, are employed to assess the models. The results indicate that the Decision Tree model consistently outperformed the other models, achieving high accuracy and F1-scores in both evaluation scenarios. The Naïve Bayes, SVM, and Neural Network models also showed reasonable performance, although slightly lower than the Decision Tree model. Based on the findings, it is recommended to utilize the Decision Tree model for diabetes diagnosis due to its high accuracy and F1-scores. However, further research and validation on larger and more diverse datasets are needed to ensure the generalizability of these results. Additionally, exploring the combination of different data mining techniques or the incorporation of additional factors, such as genetic data or lifestyle factors, could enhance the predictive capabilities of the models.
  • Item
    PREDICTING HIGHER EDUCATION STUDENTS’ PERFORMANCE USING DATA MINING TECHNIQUES: THE CASE OF ETHIOPIAN POLICE UNIVERSITY
    (Mekelle University, 2025-09-24) Abdureman Jula
    Predictive modelling for students’ performance is an innovative methodology that can be utilized by higher education institutions. Accurate and reliable student performance forecasts in higher education are crucial in order to minimize inefficient utilization of resources and funds at universities. The objective of this research is to develop a predictive model for higher education students’ performance using data mining techniques. The study followed the six-step hybrid methodology of the Knowledge Discovery Process model, such as understanding of the problem domain, understanding of the data, preparation of the data, data mining, and evaluation of the discovered knowledge and use of the discovered knowledge to achieve the goal. The study tries to understand factors affecting higher education student performance based on the data collected from Ethiopian Police University from the years 2008 up to 2012 E.C. After data preparation using data cleaning, classification algorithms such as J48 Decision Tree, PART Rule induction, Nave Bayes, Logistic regression, Support Vector Machines, and Multilayer Perception Neural Network were used for all experiments due to their popularity in recent related works. The study used a dataset containing 5254 instances, 18 attributes, and one outcome variable to run the experiments. The WEKA 3.9.5 open source software was used as a data mining tool to implement the experiments. The study also used a 10-fold cross validation and 66% split test modes for splitting the data into training and test datasets. The result of the study showed that the J48 decision tree algorithm has registered the best classification accuracy of 98.4%. The results obtained in this study are interesting and encourage the design of a model that predicts higher institution students' performance. The major factors affecting the students' performance were identified as previous study GPA, previous study field, program type, students’ performance, batch, financial source, and students’ status. In this study, an attempt was made to show the use of knowledge extracted by data mining. In the future, we recommend the automatic integration of data mining with a knowledge system so as to design an intelligent system.