NasNet Model for Image-Based Essay Scoring Deep Learning

Automated essay grading systems have increasingly become a focal point in the educational landscape, primarily due to their potential to enhance the efficiency and objectivity of essay evaluation processes. Recognizing this potential, the paper presents an innovative strategy for automated essay grading that leverages the NasNetMobile architecture as its foundational model. The methodology outlined involves a meticulous data division according to specific question numbers, followed by a manual aggregation of scores assigned to each question to derive a final grade for every essay. Through rigorous exper-imentation, the paper identifies an optimal data partitioning strategy, revealing that a 60% split for testing coupled with a 40% allocation for training purposes culminates in a remarkable accuracy rate of 84%. The technical implementation of this system is executed using the Python programming language, TensorFlow Keras for machine learning model development, and Google Colab for project collaboration and resource sharing. The choice of the Adam optimizer, characterized by a learning rate of 0.002 and a batch size of 32, significantly contributed to the system's adeptness in precisely classifying students' essay scores. Demonstrated through these findings, our proposed automated essay grading system showcases its substantial capability to serve as an effective and reliable tool in assessing student essays, thereby holding considerable promise for future adoption and further development within educational settings.