Learn, Predict and Reschedule: Toward a Data-Driven Crowdfunding Platform Management

As the crowdfunding industry grows increasingly competitive, platform managers lack systematic, datadriven tools to move beyond ad hoc decision-making toward an informed management of their platforms. We propose a CFP management framework that leverages Machine Learning (ML) and data analytics to enhance the aggregate success rate. In particular, we provide managers with an explainable Artificial Intelligence (xAI) tool that links success to key parameters of the platform's state. We compare several ML algorithms and combine correlation analysis (SHAP) with causality testing (Granger) to identify the most influential and actionable parameters. The management framework builds on this xAI analysis for controlling the success rate: a predictor forecasts the CFP state while a controller proposes an optimized campaign launch schedule to increase the aggregate success rate. We show the performance of a controller based on a Genetic Algorithm (GA) on a large-scale simulator built on real platform data. Our results illustrate how explainable and actionable AI enable the platform manager to understand the dynamics of its platform and control it for a better performance.