Yazar "Al-Shaikhli, Abdullah Raed Fadhil" seçeneğine göre listele

Listeleniyor 1 - 1 / 1
  • Küçük Resim
    Öğe
    ESTIMATION OF STUDENTS’ PERFORMANCE IN DISTANCE EDUCATION USING ENSEMBLE-BASED MACHINE LEARNING
    (2023-01) Al-Shaikhli, Abdullah Raed Fadhil
    Machine learning techniques applied in the educational context can reveal hidden knowledge and patterns to assist decision-making processes to improve the educational system. In recent years, predicting student success in the academic sector has increased interest in improving the shortcomings of academics and providing support to future students. Machine learning techniques have been used to build prediction models using students' academic past records to assist in this task. The performance of students in academic institutions indicates how much work such institutions need to continue to do to improve their low or even moderate performance. The importance of using machine learning techniques to utilize students' historical data to predict unknown or future performance was an important parameter that encouraged us to build the model . Due to its high generalization performance, ensemble learning has attracted great interest. The main challenges of building a strong ensemble are to train a variety of accurate base classifiers and combine them efficiently. The ensemble margin is calculated by taking the vote difference. The number of votes received by the correct class and the number of votes received by another class is commonly used to describe the success of ensemble learning. The classification confidence of the base classifiers is not considered in this formulation of the ensemble margin. In this study, we applied an ensemble classifier as a classification strategy to predict the substitute achievement prediction model based on machine learning. This model uses discrete datasets to reflect the student's interaction with the teaching model. Various classifiers such as logistic regression, naive bayes tree, artificial neural network, support vector system, decision tree, random forest and k-nearest neighbor are used to evaluate the prediction model of a substitute. Furthermore, cluster processes have been used to improve the appearance of these classifiers. We have used Boosting, Bagging and Voting Algorithms, which are the most common strategies used in the research. As a result, successful results have been obtained using ensemble approaches, and the robustness of the proposed model has been demonstrated.