Collaborative Cognition for Commodity Price Prediction

Abstract

Predicting prices for commodities is a long-researched problem. We present novel approach for predicting prices for the commodities leveraging Collaborative Cognition (CC). This approach falls in the domain of agent based models where different agents produce predictions for a certain commodity leveraging their own expertise of specific domain. Further, these agents collaborate with others to gain knowledge (full or partial) about the features used by other agents for producing predictions for the commodity prices. An agent's prediction performance may vary due to changes in commodity ecosystem and external influencing events. Since predictions are generated at both individual agent level and at group level, the group's collective performance is comparatively robust and better than individual agents' performance. For example,agent that produces predictions based on market dynamics of products may produce more accurate predictions when there is scarcity of the commodity as compared another agent that produces predictions using a time series based model. CC has been observed to be a very powerful approach for prediction problems as compared to state of art prediction methodologies.