PREDICT THE SUCCESS OF CLEAN ENERGY STARTUPS USING A MACHINE LEARNING APPROACH

Abstract

This thesis, conducted within the Master of Science program in Entrepreneurship and Innovation at BI Norwegian Business School, explores the application of machine learning, specifically the Extreme Gradient Boosting (XGBoost) algorithm, in predicting the success of clean energy startups. This study is driven by the pressing global need to meet 'net zero' emissions targets by 2050, emphasizing the crucial role of innovative clean energy solutions.

Central to this thesis is the research question: "Using a machine learning model, can founders and investors focus on specific factors to ensure the success of clean energy startups?" This is supported by two sub-questions that guide the methodology and findings. First, "How well can a machine learning model predict clean energy startups' success?" Here, the XGBoost model exhibits robust predictive performance, accurately identifying types of companies within the dataset—71% for Mergers and Acquisitions, 82% for Initial Public Offerings, 76% for Failures, and 93% for Remain Private. Second, "Can founders and investors focus on specific factors to ensure the success of clean energy startups?" In response, SHapley Additive exPlanations (SHAP) values provide detailed insights on two levels: global and local. Globally, SHAP values identify critical factors influencing the success across the sector, such as the presence of former and active investors, and the growth rates of social media sites. Locally, SHAP values offer nuanced insights into individual startup scenarios, enabling tailored strategic approaches.

By applying an advanced machine learning technique alongside strategic business analysis, this thesis aims to contribute to the sustainable energy discourse, suggesting potential frameworks that could enhance the viability and success of clean energy startups.