Abstract: The inverse problem of choice is formulated as a problem of the settings multiattribute utility model for
known-rank alternatives. This problem has the analogy with neural network learning. There is the difference
between multiattribute utility model and neural network learning: useful function parameters are subjects for
updating instead weight inputs of neurons. It is formulated the main condition alternatives reordering. They must
be not comparable in Pareto-dominance analyses. Two alternatives are changed their places due to useful
function ratio under its parameters variation. The algorithm is proposed to train the choice model. Algorithm is
illustrated by an example.
Keywords: multiattribute utility, inverse problem of choice, utility function, training model, aggregate objective
functions.
ACM Classification Keywords: H.4.2 Information Systems Applications: Types of Systems decision support
Link:
NEURAL NETWORK APPROACH TO THE FORMATION MODELS OF
MULTIATTRIBUTE UTILITY
Stanislav Mikoni
http://www.foibg.com/ijima/vol03/ijima03-01-p01.pdf
Wiki
(1.00/10)
