Early Placement Prediction in Higher Education Institutions Using Machine Learning

Namita Satpute, Dr. Ayesha Siddiqui

Abstract: In recent times, campus placement prediction has emerged as an important area of research owing to its direct influence on the future careers of students as well as on the success of educational institutes. The conventional approaches employed for such predictions are restricted in nature and are unable to derive any insights. For this reason, in this paper, an approach is proposed to predict the outcome of campus placement using machine learning techniques based on students' academics and other information. For this research, the dataset used has a set of important features including gender, scorecard grades from secondary and higher secondary classes, degree percentage, work experience, specialization, and MBA marks. Preprocessing steps such as removing unwanted features, handling missing values, and encoding categorical features are taken to prepare the dataset. Furthermore, feature engineering has been done using the creation of some new features such as total marks and average marks. Data is split into training and testing data with a split ratio of 70:30. Three machine learning algorithms, Gradient Boosting, LightGBM, and CatBoost, are built and compared based on their performance. Based on the experimental findings, it is observed that LightGBM performs better with a maximum accuracy of 86%, whereas Gradient Boosting and CatBoost produce the same accuracy rate of 80%. In order to ensure better understanding, various visualization techniques, including confusion matrix, heat maps, and accuracy charts, have been applied. As seen from the above discussion, it can be concluded that the suggested method proves successful in utilizing machine learning for prediction purposes.

Keywords: Machine Learning, Student Placement Prediction, LightGBM, Gradient Boosting, CatBoost, Feature Engineering, Classification, Educational Data Mining