Predicting the success of early-stage African startups using machine learning
Date
2025
Authors
Ndung'u, M. W.
Journal Title
Journal ISSN
Volume Title
Publisher
Strathmore University
Abstract
Africa's share of global venture funding is estimated to be around 1%; meaning that only a very small portion of worldwide venture capital investment goes towards African startups. This presents a challenge for entrepreneurs, investors, and policymakers seeking to foster innovation and economic growth. This study aims to bridge this gap by leveraging machine learning models to predict the success of African startups based on key factors: business operating status, number of funding rounds, and business age. Unlike prior research, which has predominantly focused on Western markets and defined success through acquisitions or IPOs, this study specifically examines African startups, addressing the continent’s unique entrepreneurial landscape. The research utilizes CrunchBase data spanning from 2000 to 2024, encompassing 28,851 startups, applying three machine learning models—Logistic Regression, Support Vector Machines, and Random Forest—to evaluate startup success. The dataset was split into training and validation sets, ensuring robust model performance assessment. Results indicate an exceptionally high accuracy of 99-100%, with strong sensitivity but lower specificity, highlighting potential dataset imbalance. Despite this, the machine learning models outperform traditional probability-based approaches by capturing non-linear relationships and complex interactions between startup success factors. This provides a more nuanced and data-driven approach to early-stage business evaluation compared to simplistic probabilistic models. The findings offer practical implications for investors by enabling more informed decision making, for entrepreneurs by identifying key success drivers, and for policymakers by informing strategies that enhance startup ecosystems in Africa. Future work should focus on balancing the dataset, incorporating additional predictive features, and expanding testing to ensure greater generalizability. This study contributes to the growing body of research on startup success prediction, offering a tailored approach for the African market and providing valuable tools for practitioners in the entrepreneurial and investment space.
Description
Full - text undergraduate research project
Keywords
Citation
Ndung’u, M. W. (2025). Predicting the success of early-stage African startups using machine learning [Strathmore University]. http://hdl.handle.net/11071/16147