Towards Collaborative Multidimensional Query Recommendation with Triadic Association Rules

This paper describes a new personalization process for decisional queries through a new approach based on triadic association rules mining. This process exploits the decision query log files of end users and follows these five steps: (1) generation of a triadic context from the multidimensional query logs of OLAP1 query analysis server; (2) mapping the triadic context into the dyadic one; (3) computation of (conventional) dyadic association rules; (4) generation of triadic association rules through a factorization process of dyadic ones and convey a richer semantics. The aim of the personalization approach which is based on triadic rules is to recommend new decision queries to OLAP end users sharing some common properties. This paper aims at helping this class of users by recommending them personalized OLAP queries that they might use in their future OLAP sessions. To validate the approach, the authors developed a software prototype called P-TRIAR (Personalization based on TRIadic Association Rules) which extracts two types of triadic association rules from decision query log files. The first type of triadic rules will serve to the recommending queries by taking the collaborative aspect of OLAP users into account. The second type of triadic rules will enrich user queries. Preliminary experiments were conducted on both real and synthetic datasets to assess the quality of the recommendations in term of precision and recall measures, as well as the performance of their on-line computation.