Enhancing Prediction of Employability of Students: Automated Machine Learning Approach

Enhancing the employability of students has recently become an important strategic goal for most institutions of higher education. With education becoming increasingly employment-oriented, a university's reputation might suffer greatly if a significant proportion of its graduates are unable to find work. Universities generally store huge amounts of student data such as student profiles, student academic records, and student behavioral records. Due to rapid scientific developments in Big Data Analytics and Machine Learning (ML), such data can be analyzed effectively to bring about great returns in predicting the employability of students. Several initiatives and research publications have developed their own ML models for predicting employability, but each has its own set of challenges and inadequacies. The primary purpose of this study is to investigate the usage of Automated Machine Learning (AutoML) as a method to increase the accuracy of prediction for the employability of students and to reduce the complexities involved in choosing the best model and corresponding hyper-parameters for a given student dataset. This research carries out AutoML using the tool Auto-Sklearn which automates the model selection and hyperparameter optimization stages of the ML pipeline. Experiments are performed where ML models are trained using Decision Tree algorithm, Gaussian Naive Bayes algorithm, Multilayer Perceptron, K-Nearest Neighbor and AutoML and the performance metrics accuracy of prediction and Matthew's Correlation Coefficient (MCC) are used to determine the best ML method and the most important employability factors. This research acknowledges that a model suitable for the dataset of one higher studies institution might not be suitable for other higher study institutions with different datasets, which is evident even in the literature for employability prediction, where, different studies corroborate different models to be the best.