We present methods to predict future market conditions and price trends from historical data, and we describe how these predictions can be used by an autonomous agent to make strategic and tactical sales decisions. The methods are based on learning dominant market conditions, such as over-supply or scarcity, from historical data and using this knowledge, together with real-time observable information, to identify the current market conditions. We use a Gaussian Mixture Model to represent the price density and a Markov process to forecast market changes and to predict price density over a planning horizon. We validate our methods by presenting experimental results in predicting price trends in the customer market for the Trading Agent Competition for Supply Chain Management.
|Title of host publication||Edited Volume of the 2nd Smart Business Network Initiative Discovery Event|
|Editors||Peter Vervest, Eric van Heck, Kenneth Preiss, Louis-Francois Pau|
|Number of pages||20|
|Publication status||Published - 2008|