Handling Qualitative and Quantitative Preferences with Constraints in Interactive Applications
Abstract
Preference elicitation is very important for interactive systems. A potential buyer
typically has speci c expectations in the attributes of the product he or she is interested
in. While the current interactive systems allow users to provide some keywords
and other information in order to lter and obtain only what they need, users feel the
product they get does not necessarily meet their satisfaction. This thesis proposes a
new interactive system that enables buyers to express their needs and desires. Users
are given the ability to elicit their requirements and preferences in a friendly and
interactive way. The system will then provide a list of suggestions satisfying user
requirements and maximizing desires. Requirements and preferences are managed
respectively as a set of hard constraints and soft constraints where the latter can be
quantitative (numerical), qualitative (ordinal), or both. This is an optimization problem
where the optimal solutions (best outcomes) are those satisfying hard constraints
and maximizing user preferences. We use the C-semiring and the CP-net to represent
the set of quantitative and qualitative preferences respectively. The branch and
bound method is then applied in order to provide the users with a list of Pareto optimal
outcomes satisfying the hard constraints and optimizing the preferences. We use
approximation techniques in order to convert conditional and qualitative preferences
to soft constraints. The proposed interactive system is enhanced with a component
that learns from other buyer preferences and makes a set of recommendations using
data mining techniques including classi cation, association rules, and cluster analysis. High- delity prototyping, an evaluation framework, and a Volere requirements speci
cation template are used in order to obtain user feedback on the interactive design
for the proposed system. For better management of constraints and preferences, the
well-known constrained CP-Net model is extended to quantitative constraints. This
extended model, named weighted constrained CP-net, takes a set of constraints and
preferences expressing user requirements and desires. It then returns a set of outcomes
provided in the form of a list of suggestions. The results of experimental tests,
conducted to evaluate the performance of the interactive system, are very promising.