CLOUD-EDGE CONTINUUM FRAMEWORK FOR ADMISSION DATA MANAGEMENT USING DEEP LEARNING MODEL
Abstract
The surge in applications necessitates a more intelligent and automated Higher Education Admission Management System (HEAMS). This research proposes a novel Deep Learning (DL)-based HEAMS utilizing a Cloud-Edge Architecture. The first step collects applicant data like transcripts, test scores, essays, and recommendations. Edge devices perform initial cleaning and preprocessing on these data to ensure quality and privacy. These preprocessed data using normalization and feature extraction using the Latent Dirichlet Allocation (LDA) are then transferred to the cloud where DL models, such as Convolutional Neural Networks (CNNs) for essays or Recurrent Neural Networks (RNNs) for transcripts, are trained. These models learn complex patterns from historical labeled data (admitted/not admitted) to predict an applicant’s success probability. During application evaluation, new data are fed through the trained models on the edge, generating probabilities for predefined classifications — high-potential, moderate, or low-potential. The cloud receives these probabilities and combines them with predefined admission criteria like minimum GPA. This combined analysis leads to a final classification using Novel Lite Convolutional Neural Network with Hybrid Leader-based Optimization (Lite CNN-HLO) for each applicant — admitted, waitlisted, or rejected and admission management system by refining admission decisions for admitted, waitlisted, and rejected applicants based on institutional priorities and constraints. The system not only generates classifications but also provides detailed model score breakdowns for transparency. This Cloud-Edge HEAMS offers improved efficiency, reduced workload for admissions staff, and potentially fairer decisions by mitigating bias through data-driven analysis.
