| Data drives all aspects of our society, from everyday life, to business, to medicine, and science. It is well-known that query personalization can be an effective technique in dealing with the data scalability challenge, primarily from the human point of view. In order to personalize their query results, user's need to express their preferences in an effective manner. There are two types of preferences: qualitative and quantitative. Each preference type has advantages and disadvantages with respect to expressiveness. The most important disadvantage of the quantitative model is that it cannot support all types of preferences while the qualitative model can only create a partial order over the data, which makes it impossible to rank all the results. The hypothesis of this dissertation is that it is possible to overcome the disadvantages of each preference type by combining both of them, in a single model, using the notion of intensity. This dissertation presents such a hybrid model and a practical system that has the ability to convert the intensity values of qualitative preferences into intensity values of quantitative preferences, without losing the qualitative information. The intensity values allow to create a total order over the tuples in the database that match a user's preferences as well as to significantly increase the coverage of preferences. Hence, the proposed model eliminates the disadvantages of the existing two types of preferences. This dissertation formalizes the hybrid model using a preference graph and proposes an algorithm for efficient preference combination, which is evaluated in an experimental prototype. The experiments show that: (1) intensity plays a crucial role in determining the order of selecting and applying the preferences, and simply ordering the preferences based on the intensity value is not necessarily sufficient; (2) the model can achieve three orders of magnitude increase in coverage compared to other alternatives; (3) the solution proposed outperforms other Top-k algorithms by being able to use both qualitative and quantitative preferences at the same time, and (4) the algorithm proposed is efficient in terms of time complexity, returning tuples ordered by the intensity value in a matter of seconds.;Keywords: Qualitative Preferences, Quantitative Preferences, Top-K Ranking. |
