Stochastic Congruence Tuning Optimized Discriminative AI (SCOTO-DAI) Model Skill Based Employability Prediction In Higher Education

Abstract

Enhancing student usability is key priority for organization of superior education, as it enables early identification of students who may be at risk of unemployment after graduation. By predicting employability outcomes in advance, institutions take timely and effective measures to support these students. This proactive approach allows management to implement targeted training and boosting their chances of securing employment. Incorporating highly developed expertise namely ML, Deep Learning, to further improve the employability prediction. But the accurate employability prediction with minimal error is a major concern, especially when dealing with large or small scale datasets. To improve the accuracy of employability prediction, a novel Stochastic Congruence Tuning Optimized Discriminative AI (SCOTO-DAI) model is developed. The proposed SCOTO-DAI model utilizes Discriminative AI model called deep transfer learning for accurate student employability prediction in the given organization. The discriminative Artificial Intelligence (AI) is an efficient method for classification and prediction based on existing data. It includes different procedure namely data acquisition, pre-processing, feature selection, classification as well as fine tuning. During the data acquisition phase, student information samples are gathered as of relevant database to facilitate accurate prediction of employability. Deep transfer learning is employed by adapting a pre-trained deep learning classification model called Convolutional Neural Learning model for employability prediction. A CNL model comprises several layers. Within the hidden layers, data preprocessing is conducted to handle the missing values and eliminate outliers. Following this, dimensionality of database is reduced through choosing most significant aspects. Finally, classification occurs at the third hidden layer (fully connected layer), where features are analyzed to distinguish the student data samples into Employable and less Employable at output layer. Followed by, Knowledge acquired from the earlier predictions made by the pre-trained method is transmitted to new method, that helps improve the efficiency of feature learning and enhances prediction accuracy. A significant step in this process is fine-tuning; Shuffling Shepherd Optimization algorithm is employed to reduce training and validation errors, thereby boosting the precision of student employability prediction. Experimental evaluation of SCOTO-DAI is carried out with different factors with number of student data.