Deep Learning Based Automatic Essay Scoring Using Convolutional Neural Network with Alexnet (CNN-A)

Essay assessments, especially in mathematics, present unique challenges due to the variation in answer formats and the potential for subjective scoring. Manual evaluation becomes increasingly inefficient in large-scale educational settings. To address these issues, this study proposes an automated scoring system for students' final answers using a deep learning approach based on the AlexNet Convolutional Neural Network (CNN) architecture. The system processes images of handwritten final answers, labeled according to their actual scores, and trains a model to classify them into pre-defined score categories. Experimental results show that the proposed model achieves an accuracy of 84.52%, outperforming baseline models such as DenseNet and 1D CNN. These findings highlight the potential of the system to support more efficient, consistent, and scalable evaluation in educational settings.