Browsing by Author "Bairamova D."

Now showing 1 - 3 of 3
  • Thumbnail Image
    ItemOpen Access
    DATA COLLECTION TO IDENTIFY STUDENTS AT RISK OF NOT COMPLETING A COURSE USING MACHINE LEARNING
    (СДУ хабаршысы - 2023, 2023) Bairamova D.
    Abstract. One of the most important methods in the study of various subjects is the understanding at an early stage of the learning process on the part of both the teacher and the student that the student is in a risk group that will not complete the course successfully. Identifying this group of students at an early stage of learning increases the level of motivation of students to start studying well in time and can help the teacher individually determine which student needs help. Before identifying a group of students at risk of not completing the course successfully, an important part is to collect and prepare the necessary data (predictors) for teaching machine learning algorithms. Currently, this is necessary for both online and offline education. In the presented method of determining a group of students, various types of algorithms were used, where one of the best results of determining a group of students with risk and without risk was shown by Logistic Regression with a high AUC =0.8003. The SMOTE method was used in the work, which coped well with the problem of data imbalance of the "Pass" and "Not Pass" classes, while increasing the accuracy of the forecast for the minority class "Not Pass" by 11%. Using certain predictors of student performance, it is possible to derive additional information such as the level of interest in the lesson. the determination of the final score for the lesson, a certain category (A, B, C, D) of students with different characteristics and other indicators that contribute to the involvement of students in the lesson at the earliest stage of learning.
  • Thumbnail Image
    ItemOpen Access
    Identification of Students at Risk of Not Completing the Course Using Machine Learning
    (2023) Bairamova D.
    This dissertation is dedicated to the topic of identifying students at risk of not successfully completing a course at an early stage of their education using machine learning algorithms. In this study, the final exam score, academic performance category, and the risk group of students failing to complete the course are determined for accurate and detailed identification of students at risk. Research shows that machine learning algorithms such as LightGBM Regressor (for the final exam score prediction), Logistic Regression (for identifying two groups of students - those who will complete and those who will not complete the course), and K-Means (for identifying the academic category) can help identify students who need assistance from teachers with high accuracy of prediction. Detecting this group of students at an early stage of their education can enhance students’ motivation for further learning and assist teachers in individually and timely identifying which student requires help.
  • Thumbnail Image
    ItemOpen Access
    PREDICTING COURSE GRADES OF STUDENTS’ ACADEMIC PERFORMANCE USING THE LIGHTGBM REGRESSOR.
    (СДУ хабаршысы - 2023, 2023) Bairamova D.
    Abstract. In the modern world, using all available opportunities and technologies, special attention should be paid to the development of the education system of students, since education serves as the basis for the development of the future generation. Nowadays, thanks to the use of available Artificial Intelligence methods, it is possible to predict various events, anomalies or other important things. With the help of machine learning, it is possible to predict at an early stage of a student's education whether he will finish the course successfully or not. In this study, it is proposed to predict the final score which student will receive at the end of the course using a number of predictors as an assessment for the first quiz and 3 types of tasks using the LightGBM regressor, which is a high-performance algorithm with gradient boosting. The results of using the LGBM regressor using GridSearchCV allowed to determine the best settings of hyperparameters from three selected tree-like boosting methods: 'dart', 'gbdt', 'goss'. The GOSS method was determined to be the best of the three methods listed with an estimate of R2 score in 0.81, which is 0.24 more than the R2 score of the Linear Regression forecast of — (0.57).