An agent-based model with social interactions for scalable probabilistic prediction of performance of a new product

Riccardo Parviero, Kristoffer H. Hellton, Ola Haug, Kenth Engø-Monsen, Hanne Rognebakke, Geoffrey Canright, Arnoldo Frigessi, Ida Scheel

Abstract: Highlights•A continuous-time, intensity-based modeling approach for agent-based models (ABMs).•ABM including both agent-to-agent interactions and external effects.•Inference for model parameters based on individual behavior in the training period.•Computationally efficient algorithms for calibration and prediction.•Exogenous sources of influence distinguished from viral, word-of-mouth spread.•Macro- and micro-validation on a real data set and what-if scenario simulations.