Context-Aware User and Service Profiling by Means of Generalized Association Rules

Abstract

Context-aware applications allow service providers to adapt their services to actual user needs, by offering them personalized services depending on their current application context. Hence, service providers are usually interested in profiling users both to increase client satisfaction, and to broaden the set of offered services.Since association rule extraction allows the identification of hidden correlations among data, its application in context-aware platforms is very attractive. However, traditional association rule extraction, driven by support and confidence constraints, may entail either (i) generating an unmanageable number of rules in case of low support thresholds, or (ii) discarding rare (infrequent) rules, even if their hidden knowledge might be relevant to the service provider. Novel approaches are needed to effectively manage different data granularities during the mining activity.This paper presents the CAS-Mine framework to efficiently discover relevant relationships between user context data and currently asked services for both user and service profiling. CAS-Mine exploits a novel and efficient algorithm to extract generalized association rules. Support driven opportunistic aggregation is exploited to exclusively generalize infrequent rules. User-provided taxonomies on different attributes (e.g., a geographic hierarchy on spatial coordinates, a temporal hierarchy, a classification of provided services), drive the rule generalization process that prevents discarding relevant but infrequent knowledge.Experiments performed on both real and synthetic datasets show the effectiveness and the efficiency of the proposed framework in mining different types of correlations between user habits and provided services.