Predicting Students Performance Using Data Mining Approach (Case Study: IT Del)

Predicting student performance stands as a critical task to comprehensively assess their academic achievements, whether they classified as the category of good, sufficient, or bad. The outcome of a student's final exam can be influenced by a multitude of factors, among which the quality of the learning process and teaching methodology hold considerable significance. Recognizing the intricate interplay of these factors, it becomes evident that the integration of a satisfaction survey becomes imperative. In the realm of this research, we harness the potential embedded within both student satisfaction survey data and their ultimate academic grades. Overcoming existing challenges by an imbalanced dataset, we employ a holistic approach by utilizing the Support Vector Machine - Synthetic Minority Oversampling Technique (SVM-SMOTE) to rectify the skewed distribution of data. To further enhance prediction capabilities, we introduced an innovative technique known as Deep Feature Synthesis. This advanced approach generating 36 novel features. The culmination of our research underscores the superiority of the random forest algorithm, especially when enriched by the harmonious merging of Deep Feature Synthesis and SVM-SMOTE through the strategic implementation of Random Search. Notably, our model achieves an impressive accuracy rate of 69%, complemented by 47% precision, 52% recall, and 47% F1-score.