Electronic Thesis/Dissertation
 

A Machine Learning Approach to Predicting Startup Success and Enhancing Investment Decision-Making

Open Access Deposited

Logistic Regression, Decision Trees, Random Forests, Gradient Boosting Machines, XGBoost, and a stacking ensemble model. The stacking ensemble model always performed better than all baselines. It demonstrated the best classification performance while still maintaining the ability for post hoc interpretability with SHAP. The results give the startup ecosystem actionable insights and further the field of AI-assisted decision-making in venture capital. Future research may improve predictive capabilities by combining unstructured data sources and incorporating qualitative factors, such as the entrepreneur's experience and the timing of the market.

Venture capital investments play a central role in driving innovation and economic growth. Yet, startups experience a failure rate exceeding 90%, with over 60% of venture capital investments yielding no returns. This ongoing lack of efficiency shows how important it is to find more precise, data-driven ways to judge the feasibility of a firm. This study develops and evaluates a machine learning–based predictive framework to classify startup success, leveraging a structured dataset that includes firm-level, financial, and geographic indicators. Key features were selected through recursive feature elimination and SHAP-based interpretability analysis. A comparative evaluation of multiple supervised learning models was conducted on the following models

Author Language Date created Type of Work License
  • All rights reserved
Rights statement GW Unit Degree Advisor Committee Member(s) Persistent URL

Notice to Authors

If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.

Thumbnail Title Date Uploaded Visibility Actions
Preview of RiveraIII_gwu_0075A_17427.pdf RiveraIII_gwu_0075A_17427.pdf 2025-12-11 Open Access