Prediction of IPO Subscription – A Logistic Regression Model

Abstract

The main objective of this research paper is to apply logistic regression to estimate IPO subscription status in terms of oversubscription or under subscription. For this purpose, we used SMOTE (Synthetic Minority Oversampling Technique) to generate minority class cases to rectify class imbalance problems and classification model logistic regression function to further classify the cases into majority class and minority class. KNIME (Konstanz Information Miner) and R Studio were used, as Integrated Development Environments (IDE), to develop the model. The results were quite encouraging with more than 90% accuracy levels for both training and testing datasets. The model was tested with different train-to-test ratios. The model and the results of the study can be used by firms and individuals involved in capital markets to predict the subscription status of a public offering. Further, there is ample scope to improvise the model by using different sets of variables and by applying different machine learning algorithms.