Predictive Analytics for Startups Success: Acquisition Prediction based on Machine Learning Techniques

Abstract

With the rise of the internet, startups number has substantially increased in the last two decades. Although many of them have succeeded to revolutionize their sectors, many of them have shut down a few months or even a few years after the foundation. In this paper, we propose a machine learning approach for predicting startups success based on historical data. Almost 840 startup data have been finely preprocessed to extract 35 features that characterize each a particular aspect of the studied startups. Afterward, computational models based on machine learning techniques were developed and tested using a cross-validation approach. The main objective is to predict the success of the startup, especially in terms of mergers and acquisitions. In particular, several models have been applied namely Artificial Neural Networks (a.k.a. ANNs), Support Vector Machines (a.k.a. SVMs), Random Forests, Bagging, Stacking, and Gradient Boosting. Overall results were very promising since the best model succeeded in the prediction with an accuracy rate of 85%. Furthermore, a feature importance study was also conducted to analyze the best predictors of a startup acquisition.