The weighted majority algorithm
Information and Computation
| Hi-index | 0.00 |
This paper extends the current Potluck Problem to include multiple goods in the relevant problem scenario, thereby bringing in the question of estimation of consumers' choice behaviors in the system. Given a variety of dishes to choose from at any dinner instance, an agent chooses a set of dishes in a particular proportion such that his utility from the total consumption is maximized. Such choice behaviors of agents determine the demand structure of the dishes offered in a potluck dinner. In this paper, we use mathematically specified models of consumer choice to represent these behaviors. These models are used to build a set of predictors that are capable of computing certain functional forms and variables (including seasonal and trend variables) of the demand system. Such predictors existing in the system are then used by agents in estimating the demand at future potluck dinner instances.