Student Placement Performance Analysis and Prediction Systems at Higher Education Institutions in South India Region: A Machine Learning Approach with Ranked Features

Abstract

Technical and non-technical industries strive to provide employment opportunities for undergraduates who have completed their degrees or have successfully passed qualifying exams for higher education. Educational institutions constantly seek improved technology to enhance operations, facilitate decision-making, and develop innovative strategies. Predicting student placement outcomes is crucial for higher education institutions. This study uses machine learning (ML) to analyze student placement performance. It focuses on higher education institutions in South India. There are 39 features in the dataset, such as academic performance, skills, and socio-economic factors. Feature importance is calculated using four metrics: information gain, gain ratio, Gini index, and chi-square. The most influential predictors identified were course percentage (F6), attendance (F5), coding, and course type (F2), validated by high x² scores (F6: x²=69.847) and Information Gain (F6: 0.13). Six ML classifiers were tested on the top 20 to 25 features. CatBoost had the best performance with an AUC of 0.984, an accuracy of 94.6%, and a precision value of 0.94. AdaBoost followed closely with an accuracy of 94.4% and F1: 0.944. The confusion matrix and AUC evaluations showed how effective the models were. Accuracy = 94.4%, F1 = 0.944. In this research paper, we have tried to explain the important things that influence getting an opportunity in a student's career. We have also created a new synthetic dataset and utilized various machine learning techniques to develop a new model to determine its accuracy and determinism. We identified significant features from the dataset and verified their accuracy, resulting in a model that performed well in all aspects.