Price Prediction in a Trading Agent Competition

Abstract

The 2002 Trading Agent Competition (TAC) presented a challenging market game in the domain of travel shopping. One of the pivotal issues in this domain is uncertainty about hotel prices, which have a significant influence on the relative cost of alternative trip schedules. Thus, virtually all participants employ some method for predicting hotel prices. We survey approaches employed in the tournament, finding that agents apply an interesting diversity of techniques, taking into account differing sources of evidence bearing on prices. Based on data provided by entrants on their agents' actual predictions in the TAC-02 finals and semifinals, we analyze the relative efficacy of these approaches. The results show that taking into account game-specific information about flight prices is a major distinguishing factor. Machine learning methods effectively induce the relationship between flight and hotel prices from game data, and a purely analytical approach based on competitive equilibrium analysis achieves equal accuracy with no historical data. Employing a new measure of prediction quality, we relate absolute accuracy to bottom-line performance in the game.

Economic reasoning and artificial intelligence

“…We can evaluate the effectiveness of economic modeling by examining agents built by AI designers for specified tasks. For instance, in a study of AI trading agents competing in a shopping game (21), an agent using standard price equilibrium models from economics (specifically, Walrasian equilibrium) achieved comparable prediction accuracy to sophisticated machine-learning approaches without using any data, even though none of the other agents employed equilibrium reasoning.…”

A procedure using support vector data description and mutual information for end price assessment in online C2C auction

“…Collecting these data through well designed spiders [6,28], cleaning the data, and extracting the relevant features by applying text mining techniques [22] are tedious tasks. Researcher have used these data for various purposes such as understanding the buyers' behavior [1], detecting anomalies such as shill bidding [9], analyzing the effect of different auction attributes on the final price [12], predicting the final price [1,4,7,12,18,22,24,25,28], suggesting bidding strategies [17], and designing bidding agents [27]. For example, Bajari and Hortacsu use a dataset of eBay coin auctions to find the determinants of bidder and seller behavior [1].…”

“…The first example is the Trading Agent Competition (TAC)-a platform for building agents to bid on airline, hotel, and ticket prices [27]. Several TAC competitors have explored a range of methods for price prediction including historical averaging, neural nets, and boosting.…”

Intelligent spider for information retrieval to support mining-based price prediction for online auctioning

“…In that competition, virtually all participating agents employ some approaches for predicting hotel prices. Furthermore, Wellman, Reeves, Lochner, and Vorobeychik (2002) developed agents with a challenging market game in the domain of travel shopping. A pivotal issue concerning travel agents is uncertainty about hotel prices, which significantly affect on the relative cost of alternative trip schedules.…”

“…The prices are finalized until the auctions are closed (Wellman et al, 2002). No useful information is revealed before the final call.…”