Major Project Report on Analysis and Prediction of Mergers & Acquisitions

Abstract

Sustaining in a highly competitive market is difficult and a challenge that the company looks to overcome by providing the best to customers than the existing options available. To diversify their operations and attain the position of the market leader, most companies opt for Mergers or acquisitions for having power or success in this changing environment. The black swan events have resulted in many Mergers, Acquisitions, and takeovers due to firms not able to adapt to changing times and the constantly innovating firms acquiring such targets which would benefit them in the longer term. In this project, we aim to study and explore the multi-class prediction problem of identifying the status of the company whether it should opt for mergers, acquisitions, IPO, or continue in operating mode. Firstly, we provide exploratory data analysis using popular data visualization tools to gain useful insights from Crunchbase and WorldBank datasets. Secondly, we propose a novel method to address the prediction problem to identify the status of the company using machine learning techniques. We have employed various under-sampling methods to deal with the problem of imbalance in the dataset. Also, we incorporate the additional factors like macroeconomic variables and Intellectual property rights of the home country which are considered useful from an M & A perspective. We have performed experiments to determine the best-performing model among machine learning techniques like Logistic Regression, K-nearest neighbor, Random Forest, and XG Boost and compare the results with the baseline modeling using appropriate evaluation metrics. The Edited Nearest Neighbour under-sampling technique presents the best results using K-nearest neighbor and closely followed by XGBoost Classifier Model. Experimental results demonstrate that the proposed approach outperforms the existing methodology adopted by the researchers in the past.