PREDICTING INVESTOR SUCCESS USING GRAPH THEORY AND MACHINE LEARNING

We extract a large dataset of venture capital financing and related startup firms from Crunchbase. This paper examines how network position determines the success rate of investors. Precision in determining which investors will be successful is relatively high, but it is in fact easier to predict unsuccessful investors. Graph-theoretic features may be used in machine-learning algorithms to improve predictions of VC performance. This study has implications for how startups and private bank investors may choose investors and suggests a two-step approach where segmentation by industry is done first, followed by community construction within industry. In short, choosing a VC should be first based on subsetting VCs who have a focus in the industry of the startup followed by the use of a machine-learning model. This cross-disciplinary paper generates insights by combining financial data with graph-theoretic ideas and machine-learning algorithms.