Full Citation

Title: Similarity-Based Recommendation of OLAP Sessions

Citation Type: Dissertation/Thesis

Publication Year: 2013

ISBN:

ISSN:

DOI:

NSFID:

PMCID:

PMID:

Abstract: This dissertation presents an approach for recommending OLAP sessions, in a collaborative filtering context, and based on similarity measures between queries and sessions.After briefly reviewing classical techniques for usage mining in Web Page Recommendation, a study of recommender systems in Databases and Data Warehouses allows to identifyseveral shortcomings. Indeed, sequential aspects are rarely addressed in these works and no approach ever considered to recommend sessions. Besides, queries are rarely synthesized for the recommendation and are often chosen among past queries. This dissertation answers these shortcomings by proposing a set of requirements to take into account in arecommendation context. Since the recommender system is based on a similarity measures, a study of classical measures in information retrieval is also presented. Afterward several similarity measures are extended in an OLAP context and are organized in a three-level approaches between OLAP logs. Similarity measures between logs depend on similarity measures between sessions that depend on similarity measures between queries. Then, a recommender system based on similarity measure between sessions is proposed. Three phases compose this system. The first phase aligns the log sessions with the current session and identifies possible recommendations. The second phase ranks each recommendation by identifying densest areas of similar queries in the log sessions. The last phase adapts the recommendation, ranked first to the current session, using patterns extracted from the log and the current session, and recommends it. Also, the recommender system is assessedin terms of efficiency and effectiveness with sessions coming from synthetic log generations or logs whose sessions have been devised by Masters students in Business Intelligence.

User Submitted?: No

Authors: Aligon, Julien

Institution: Universitrans Rabelais de Tours

Department: Laboratoire D

Advisor: Arnaud Giacometti

Degree: Ph.D

Publisher Location: Tours, France

Pages:

Data Collections: IPUMS USA

Topics: Other

Countries: