2216. Fast Estimation Process for Random Preference Model
Invited abstract in session MB-8: Preference Learning 1, stream Multiple Criteria Decision Aiding.
Monday, 10:30-12:00Room: Clarendon SR 2.08
Authors (first author is the speaker)
| 1. | Moha Ghaderi
|
| Economics and Business, Pompeu Fabra University |
Abstract
Random Preference Model (RPM) is a nonparametric and flexible choice model capable of capturing a wide range of context-dependencies. In its most general case, it takes the population of permutations (complete rankings) over the set of choice options as the primitive and constructs: 1) a probability distribution over this set, and 2) a probability distribution over preference parameters (e.g., utility functions) that yield each specific permutation. Therefore the space of parameters grows exponentially with the size of choice options, making the estimation process intractable. Approximation algorithms rely restricting the RPM support, using partial instead of full rankings, or a combination of both. In all these case, however, RPM still requires sampling from the utility space to approximate the probability distribution over the preference parameters space, making it computationally expensive. I introduce an fast estimation process based on expectation-maximization algorithm that circumvents the need for a sampling and thus substantially reducing the estimation cost. I compare the proposed method with the sampling-based estimation using simulated choice data.
Keywords
- Behavioural OR
- Decision Analysis
- Algorithms
Status: accepted
Back to the list of papers