Startup Success Prediction with PCA-Enhanced Machine Learning Models

Abstract

This study evaluates the effectiveness of various machine learning algorithms in predicting startup success and explores the performance improvement achieved by applying Principal Component Analysis (PCA) to the models. By analyzing logistic regression, support vector classifier (SVC), XGBoost, and other supervised learning algorithms, the study demonstrates that PCA enhances the generalization performance of most models. Notably, Support Vector Classifier (SVC) showed an accuracy of 0.78, precision of 0.83, recall of 0.73, and F1 score of 0.74 without PCA, but performance significantly improved with PCA, recording an accuracy of 0.90, precision of 0.90, recall of 0.89, and F1 score of 0.89. Academically, this research contributes to the literature by examining how dimension reduction can boost the accuracy of machine learning models for startup success prediction, providing a valuable intersection of machine learning and venture capital studies. Practically, it offers investors AI-driven decision- making tools to enhance the precision of investment evaluations and better identify startups with high growth potential. Despite its contributions, this study is limited by the specific dataset used, suggesting that future research could explore various datasets and alternative dimension reduction techniques. Future studies could also assess real-time data application and incorporate deep learning models to improve predictive performance in startup success evaluation.