|
Analysis and prediction of undergraduate student dropouts using machine learning |
|---|---|
| รหัสดีโอไอ | |
| Creator | Jakkapan Satmunee |
| Title | Analysis and prediction of undergraduate student dropouts using machine learning |
| Contributor | Ekachai Naowanich, Thanaporn Patikorn |
| Publisher | Mahasarakham University |
| Publication Year | 2569 |
| Journal Title | Journal of Science and Technology Mahasarakham University |
| Journal Vol. | 45 |
| Journal No. | 2 |
| Page no. | 231-242 |
| Keyword | Analysis and prediction, machine learning |
| URL Website | https://li01.tci-thaijo.org/index.php/scimsujournal |
| Website title | Journal of Science and Technology Mahasarakham University |
| ISSN | 1686-9664 (Print), 2586-9795(Online) |
| Abstract | This study presents an analysis and comparison of the performance of eight machine learning models, namelyDecision Tree, Random Forest, Logistic Regression, K-Nearest Neighbors, AdaBoost, Hist Gradient Boosting, XGBoost, and Light Gradient Boosting, in identifying at-risk undergraduate students in Thailand. This study focuses on analyzing academic data from university students to understand the factors influencing academic success or failure, and developing predictive models to help instructors and administrators identify learning trends and implement effective teaching strategies. The dataset consists of records from 2,104 undergraduate students in science and technology from 2020 to 2024 in Thailand, with 36 data features. During data preparation, missing values were imputed using the k-nearest neighbors (k=5) method, and class imbalance was corrected using the synthetic minority sampling technique (SMOTE). Recursive feature elimination (RFE) was used to identify important features. The model performance was evaluated using cross-validation using four main metrics: accuracy, precision, recall, and F1 score. The results indicate that the Random Forest model performs the best. It achieved an accuracy of 84.32%, a precision of 0.85, a recall of 0.84, and an F1 score of 0.85. The findings suggest the significant potential of this model in the educational system, particularly in providing targeted interventions for at-risk students to reduce dropout rates and increase academic success. Future research should explore additional influencing factors and investigate other machine learning models that may provide improved performance, as well as compare the results across diverse student populations to ensure reliability and broader applicability. |
