Web Prediction Mechanism for User Personalized Search

Abstract

Personalized net search (PNS) has incontestible its effectiveness in up the standard of assorted search services on the web. However, evidences show that users� reluctance to disclose their non-public data throughout search has become a serious barrier for the wide proliferation of PNS. We have a tendency to study privacy protection in PNS applications that model user preferences as graded user profiles. We have a tendency to propose a PNS framework referred to as UPS (User customizable Privacy-preserving Search) that may adaptively generalize profiles by queries whereas respecting user specified privacy necessities. Our runtime generalization aims at hanging a balance between 2 prognostic metrics that assess the utility of personalization and therefore the privacy risk of exposing the generalized profile. We have a tendency to gift 2 greedy algorithms, specifically GreedyDP and GreedyIL, for runtime generalization. We have a tendency to additionally give a web prediction mechanism for deciding whether or not personalizing a question is useful. intensive experiments demonstrate the effectiveness of our framework. The experimental results additionally reveal that GreedyIL considerably outperforms GreedyDP in terms of potency.