This paper presents a set of guidelines, imported from the field of forecasting, that can help social simulation and, more specifically, agent-based modelling practitioners to improve the predictive performance and the robustness of their models. The presentation starts with a discussion on the current debate on prediction in social processes, followed by an overview of the recent experience and lessons learnt from the field of forecasting. This is the basis to define standard practices when developing agent-based models under the perspective of forecasting experimentation. In this context, the guidelines are structured in six categories that correspond to key issues that should be taken into account when building a predictor agent-based model: the modelling process, the data adequacy, the space of solutions, the expert involvement, the validation, and the dissemination and replication. The application of these guidelines is illustrated with an existing agent-based model. We conclude by tackling some intrinsic difficulties that agent-based modelling often faces when dealing with prediction models.
