Agent-Mediated Electronic Commerce. Designing Trading Strategies and Mechanisms for Electronic Markets

In many auction settings we find a self-interested information broker, that can potentially disambiguate the uncertainty associated with the common value of the auctioned item (e.g., the true condition of an auctioned car, the sales forecast for a company offered for sale). This paper extends prior work, that has considered mostly the information pricing question in this archetypal three-ply bidders-auctioneer-information broker model, by enabling the information broker a richer strategic behavior in the form of anonymously eliminating some of the uncertainty associated with the common value, for free. The analysis of the augmented model enables illustrating two somehow non-intuitive phenomena in such settings: (a) the information broker indeed may benefit from disclosing for free some of the information she wishes to sell, even though this seemingly reduces the uncertainty her service aims to disambiguate; and (b) the information broker may benefit from publishing the free information to the general public rather than just to the auctioneer, hence preventing the edge from the latter, even if she is the only prospective customer of the service. While the extraction of the information broker’s optimal strategy is computationally hard, we propose two heuristics that rely on the variance between the different values, as means for generating potential solutions that are highly efficient. The importance of the results is primarily in providing information brokers with a new paradigm for improving their expected profit in auction settings. The new paradigm is also demonstrated to result, in some cases, in a greater social welfare, hence can be of much interest to market designers as well.

Crowdsourcing marketplaces link large populations of workers to an even larger number of tasks. Thus, it is necessary to have mechanisms for matching workers with interesting and suitable tasks. Earlier work has addressed the problem of finding optimal workers for a given set of tasks. However, workers also have preferences and will stay with a platform only if it gives them interesting tasks. We therefore analyze several matching mechanisms that take into account workers’ preferences as well. We propose that the workers pay premiums to get preferred matches and auction-based models where preferences are expressed through variations of the payment for a task. We analyze the properties of two matching different mechanisms: Split Dynamic VCG (SDV) and e-Auction. We compare both the mechanisms with Arrival Priority Serial Dictatorship (APSD) empirically for efficiency.

The Power TAC is a competition-based simulation of an electricity market. The goal of the competition is to test retailer (broker) strategies in a competitive environment. Participants create broker agents that trade electricity. In this paper we describe our broker, which we created as a participant of the 2014 Power TAC competition. We describe the strategies for two main components of the game: the tariff market and the wholesale market. We also discuss the performance of our broker in the competition, where we were second in the final ranking.

Motivated by the energy domain, we examine a risk-averse buyer that has to purchase a fixed quantity of a continuous good. The buyer has two opportunities to buy: now or later. The buyer can spread the quantity over the two timeslots in any way, as long as the total quantity remains the same. The current price is known, but the future price is not. It is well known that risk neutral buyers purchase in whichever timeslot they expect to be the cheapest, regardless of the uncertainty of the future price. Research suggests, however, that most people may in fact be risk-averse. If the future price is expected to be lower than the current price, but very uncertain, then they may prefer to purchase in the present, or spread the quantity over both timeslots. We describe a formal model with a uniform price distribution and a piecewise linear risk aversion function. We provide a theorem that states the optimal behavior as a closed-form expression, and we give a proof of this theorem.

The Power TAC simulation emphasizes the strategic problems that broker agents face in managing the economics of a smart grid. The brokers must make trades in multiple markets and, to be successful, brokers must make many good predictions about future supply, demand, and prices in the wholesale and tariff markets. In this paper, we investigate the feasibility of using learning strategies to improve the performance of our broker, SPOT. Specifically, we investigate the use of decision trees and neural networks to predict the clearing price in the wholesale market and the use of reinforcement learning to learn good strategies for pricing our tariffs in the tariff market. Our preliminary results show that our learning strategies are promising ways to improve the performance of the agent for future competitions.

The Power Trading Agent Competition (Power TAC) is a feature-rich simulation that simulates an energy market in a smart grid, where software brokers can buy energy in wholesale markets and sell energy in tariff markets to consumers. Successful brokers can maximize their profits by buying energy at low prices in the wholesale market and selling them at high prices to the consumers. However, this requires that the brokers have accurate predictions of the energy consumption of consumers so that they do not end up having excess energy or insufficient energy in the marketplace. In this paper, we conduct a preliminary investigation that uses standard off-the-shelf machine learning techniques to cluster and predict the consumption of a restricted set of consumers. Our results show that a combination of the popular k-means, k-medoids, and DBSCAN clustering algorithm together with an autoregressive lag model can predict, reasonably accurately, the consumption of consumers.

In this paper, we present a genetic algorithmic approach to automated auction mechanism design in the context of cat games. This is a follow-up to one piece of our prior work in the domain, the reinforcement learning-based grey-box approach [14]. Our experiments show that given the same search space the grey-box approach is able to produce better auction mechanisms than the genetic algorithmic approach. The comparison can also shed light on the design and evaluation of similar search solutions to other domain problems.

The future smart grid will bring new actors such as local producers, storage capacities and interruptible consumers to the existing electricity grid along with the challenge of sustainability. Intermediary power actors, i.e., brokers, will take the burden of financial management, during the integration of these customers. This paper describes the mathematical modelling, formalization and the design of decision making systems of a winner broker agent, AgentUDE14, which competed in Power Trading Agent Competition 2014 Final (Power TAC). In this work, we divide the main trading problem into sub problems and then formalize and solve them individually to reduce the mathematical complexity. In the wholesale market, we propose a dynamic programming approach whereas our retailer algorithm uses an aggressive tariff publication policy, which exploits tariff fees, such as early withdrawal penalty and bonus payment. We show the results that AgentUDE14 is a successful agent in many metrics, analyzing the tournament data from Power TAC 2014 Finals.