Name: Kavita Dahiya
Country: India

Year: 2025
Volume: Volume-12, Issue-2 (July-December)
Page Number: 406-414
DOI: https://doi.org/10.5281/zenodo.21932971

ABSTRACT
General Total Early-Stage Entrepreneurial Activity (TEA) was of 53 economies according to the Global Entrepreneurship Monitor Report 2025/26, but only 25-30 percent of startups are carried through to full-fledged businesses, even as 84 percent of startups have social impact ambitions. The lack of AI readiness and survival issues persists, requiring new solutions based on Data Science. This paper suggests detailed frameworks that would use ensemble MLs for the multidimensional dataset of GEM: Cox proportional hazards of survival prediction and Random Forest of gap analysis. India (fourth in National Entrepreneurship Context Index in the world) confirms approach-24 per cent of the three-year survival is constricted by regional AI inequalities. Using an Indian venture sample of 2,139 (fintech: n = 892; health-tech: n = 1,247), exit-risk real-time analytics forecasting and AI-readiness signaling helped to raise survival by 35 percent. Tier-2/3 cities: survival can increase by 42 percent; low-readiness countries: improvement can take place by 29 percent. Technology-neutral models aligned to ICIDS 2026 provide scalable solutions to such barriers that can be identified by policymakers/incubators to ensure equitable entrepreneurial growth.

Keywords: GEM 2025/26, entrepreneurial survival, AI ecosystems, Cox hazards, NECI optimization