Machine Learning-based Start-up Company Lifespan Prediction: the Chinese Market as an Example

Start-ups have emerged as key drivers of economic growth, fostering innovation, job creation, and knowledge dissemination across various industries. Accurately forecasting start-up life spans is critical for investors, policymakers, and entrepreneurs to make informed decisions and optimize resource allocation. However, existing predictive models, such as linear regression and survival analysis, face challenges in capturing the complex interactions and dynamic nature of factors influencing start-up success. This paper proposes applying the XGBoost algorithm, an advanced machine learning technique, to enhance the accuracy and reliability of start-up life span predictions. XGBoost offers several advantages over traditional methods, including adaptability to various data types, robustness to outliers, and efficient computational performance. By incorporating a wide range of features, such as financial, organizational, and death reasons, the algorithm can effectively capture the complex relationships among these factors without explicit feature engineering. Moreover, applying SHAP values provides an additional layer of interpretability, aiding stakeholders in better understanding the factors driving start-up life span. Utilizing the IT Orange dataset, we investigate the determinants of startup life spans, offering valuable insights for stakeholders in the entrepreneurial ecosystem.