Human-machine hybrid prediction market: A promising sales forecasting solution for E-commerce enterprises

Abstract

Integrating machine models and humans in the prediction market could generate hybrid intelligence for sales forecasting which is of great significance to e-commerce enterprises. Following the design science paradigm, we propose a framework of human–machine hybrid prediction market and evaluate its effectiveness with economic analytical models and numerical simulations. We obtain four interesting design guidelines by modeling and analyzing three kinds of hybrid markets: the collaboration market, competition market, and co-competition market. First, machines’ collaboration behavior, i.e., prediction result sharing, does not always benefit the prediction market. Second, compared with the collaboration pattern, introducing machines as competitors with humans is a better approach. Third, co-competition between machines and humans will bring positive and negative effects simultaneously, and this kind of interaction will contribute to the prediction performance when positive effects dominate negative effects. Fourth, when human traders put correct trust on machine models, the co-competition market performs best.