Comparison of Machine Learning Models in Determining the Success of Startups

This paper is a systematic literature search that compares machine learning-based methods used to determine the success rate of startups. Startups are the driving force of the economy, yet most startups fail. This is why analysis in this area is so desirable and interesting, yet still not well researched. Using a systems approach based on structured data collection, this study describes and compares different models for assessing the success of start-ups. A literature search was conducted on journal articles from the Web of Science and Scopus databases. Finally, the paper compares the three most interesting studies based on two comparative metrics: accuracy and the “Area under the ROC (receiver operating characteristic curve) Curve” indicator. The comparison resulted in the evaluation of the best performing model for predicting the success of startups, which was Gradient Boosting.