Implementing Machine Learning Algorithms in the Screening Process of Serial Acquirers

Sammanfattning

This study investigates how machine learning algorithms can be applied to automate and enhance the efficiency of the screening process for serial acquirers. In this business model, sustained growth depends on continuously acquiring profitable companies, making the screening phase a critical component of operations. Given the complexity of fully evaluating acquisition targets, this research focuses specifically on the early-stage screening, using both financial, and non-financial, metrics to identify potentially interesting acquisition targets. At this stage, it is essential to minimize false negatives—companies incorrectly deemed uninteresting by the model—as these would otherwise go undetected by the thesis partner. Three machine learning models were developed and evaluated: a neural network, Histogram-based Gradient Boosting, and XG-Boost. All showed similar performance, achieving 81% accuracy, with 97–99% of positive examples being correctly classified as one of the two positive classes. In conclusion, the study demonstrates significant potential for applying machine learning to the screening process of serial acquirers. However, challenges remain—particularly the limited availability of data in certain regions. The methodology presented here offers a foundation for future research, which would benefit from a more extensive dataset.