AI-Driven Predictive Analytics for Business Expansion in the U.S. Start-Up Ecosystem

Authors

Abstract

This study develops AI-driven predictive models to analyze state-level start-up dynamics in the United States, leveraging historical data on firm entries, exits, job creation, and job destruction from 2015 to 2024. Using XGBoost as the pri-mary algorithm and logistic regression as a baseline, the models forecast business expansion patterns, identify high-growth regions, and evaluate ecosystem sustain-ability. Key features include venture capital incentives, labor market trends, and regional economic indicators, integrated through robust ETL pipelines. XGBoost achieved 87% accuracy in classifying high-potential states, with an F1-score of 0.88, significantly outperforming logistic regression (72% accuracy, F1-score 0.75), as evidenced by classification reports, confusion matrices, and scatter plots of predicted versus actual growth scores. Validation via 5-fold cross-validation, paired t-tests (p ¡ 0.05), and RMSE (0.12–0.15) confirms model reliability. Case studies demonstrate practical impact: a Series C-funded AI firm reduced labor costs by 35% and secured $3.2 million in incentives by relocating to Colum-bus, Ohio, while AJE Group’s AWS migration cut ETL processing time by 35%. Findings reveal Ohio, Texas, and North Carolina as emerging hubs, driven by strong public-private partnerships. The research bridges gaps in regional pre-dictive analytics, offering policymakers evidencebased incentive strategies and entrepreneurs scalable tools for market entry. Future work should incorporate real-time data and extend to non-tech sectors for broader applicability

Keywords

Submission Status