Predicting New Venture Gestation Outcomes With Machine Learning Methods
研究用多种机器学习方法预测新企业在头五年内成立或放弃的概率，发现极端梯度提升模型整体表现最佳，神经网络能准确识别早期未放弃的创业者，并揭示创业活动数量与节奏对成功的重要性。
This study explores the use of machine learning methods to forecast the likelihood of firm birth and firm abandonment during the first five years of a new business gestation. The predictability of traditional logistic regression is compared with several machine learning methods, including logistic regression, k-nearest neighbors, random forest, extreme gradient boosting, support vector machines, and artificial neural networks. While extreme gradient boosting shows the best overall model performance, neural networks provide good results by correctly classifying entrepreneurs who have not abandoned their business venture in the early stage of the gestation process. In addition, this study provides valuable insights in relation to the start-up activities leading to firm emergence. Entrepreneurs who perform a greater number of activities and who can orchestrate them at the right rate, concentration, and time are more likely to successfully launch a new business venture.