All user queries are stored in a database log. In operational environments, queries are predefined and thus the log contains the same queries where only the parameters vary. Oppositely, in analytical environments, each query addresses a given problem. Since different problems appear every day, so queries do.

However, patterns can also be found depending on the user profile. Thus, analysing those queries in an analysis log can show analysis trends and help on finding interesting analysis issues (i.e., facts), or viewpoints (i.e., analysis dimensions) that a given user could not imagine. For example, if we detect that users usually analyse production depending on the month just after detecting that sales increased in the latter semester, when a new user visualises such a sales increment, we can automatically generate the analysis of the production per month (or, at least, suggest it).

Data mining, as well as statistical techniques (whose cost could be afordable thanks to cloud and parallelism) should be used to extract profiles from user data, as well as access patterns from the query logs. Patterns found in the queries can be based on the schema of the output cube or on its data and the information it provides.

Main Advisor at Universitat Politècnica de Catalunya (UPC)
Co-advisor at Aalborg Universitet (AAU)