Preference Detection in Multi-Attribute Multi-item Choice Environments
Abstract—We propose a novel method for evaluating, detecting, and inferring preferences in choice situations involving items with multiple attributes. Our method locates characteristics that reflect the relative weight that a user gives to varying attributes belonging to an item, when these attributes are combined into a unified multi-attribute utility function. Our method enables the attainment of coefficients that reflect the preferential prism of a user in relation to items with multiple attributes. Broadly, we translate the form of a u-function into an inequality with scalar variables which define half-spaces on a plane. These half-spaces intersect and form closed shapes (in a k-dimensional world). The closed shapes with the most intersections are the most likely areas in which the vector (x1,...,xk) lies. Attaining the values of the various xi allow a computational system to restore the u-function. This enables the system to predict the alternative item’s total utilities in yet unmet choice-making scenarios. A novel extension of methods relates to the identification of inconsistent choice-making. In addressing the latter problem, we note that hyper-planes split the n-dimensional world into parts. We relate to every one of these parts (segments or rays in 1D, shapes, bounded or unbounded, in 2D), and count the number of half-spaces that contain it; this number reflects the probability that the actual (unknown) parameters are in it. Counting the number of half-spaces containing each segment allows us to consider multiple user profiles and considerations. This paves the way to the construction of more complex frameworks for understanding user choice as a multi-criteria decision making problem.
Index Terms—preference detection, multi-attribute decision making, multi-item choice, recommender systems.
Cite: Amir Konigsberg and Ron Asherov, "Preference Detection in Multi-Attribute Multi-item Choice Environments," Jounal of Automation and Control Engineering, Vol. 3, No. 1, pp. 46-50, February, 2015. doi: 10.12720/joace.3.1.46-50