Deep Learning Model Predicts College Student Academic Performance

A new deep learning model integrates multiple types of student data to predict academic performance. The approach combines grades, attendance records, learning management system interactions, psychometric surveys, and demographic information. Researchers collected data from 2,000 students and applied preprocessing steps including K Nearest Neighbor Imputation for missing values, outlier removal, and normalization.

Principal Component Analysis reduced the dimensionality of the dataset while preserving key features.

The model uses a Gated Long Short-Term Memory Unit optimized with the Dove algorithm. This combination captures time-dependent patterns in student engagement and improves convergence speed during training. Experiments conducted in Python 3.10 showed the optimized model reached 98.85 percent classification accuracy.

It produced lower error rates than several baseline machine learning methods on the same dataset.

Visualization of model outputs confirmed accurate forecasting of student performance trends. The approach provides interpretable insights into factors that influence academic outcomes. The model supports early identification of students who may need additional support. It offers a scalable method for data-driven academic performance management using publicly available datasets.