Event Shocks in Investment Networks: A Contextual-Temporal Aware Neural Point Process Framework for Startup Success Prediction

This study develops a Contextual-Temporal Aware Neural Point Process Framework (CTA-NPP), which models event shocks such as fundraising and acquisitions in a dynamic venture capital ecosystem, for predicting startup success. CTA-NPP reframes discrete organizational milestones within the ecosystem as relational and sequential events, and demonstrates how the influence of these events is modulated by the network context where they originate and propagate. CTA-NPP integrates both intra-event contextual feature interactions and inter-event temporal interdependencies to model network event shocks, which addresses the challenges of modeling how the continuously evolving investment network exerts influence on individual startups from a microscopic perspective. Empirical validation using Crunchbase data demonstrates the superiority of CTA-NPP compared to state-of-the-art benchmarks, achieving improvements in 3- and 5-year success predictions. The predictive framework offers managerial insights for venture capitalists with our data-driven capital allocation strategies that achieve up to 47% higher investment success rates than real-world investment decisions.