Abstract
Spatial data warehouses store a large amount of historized and aggregated data. They are usually exploited by Spatial OLAP (SOLAP) systems to extract relevant information. Extracting such information may be complex and difficult. The user might ignore what part of the warehouse contains the relevant information and what the next query should be. On the other hand, recommendation systems aim to help users to retrieve relevant information according to their preferences and analytical objectives. Hence, developing a SOLAP recommendation system would enhance spatial data warehouses exploitation. This paper proposes a SOLAP recommendation approach that aims to help users better exploit spatial data warehouses and retrieve relevant information by recommending personalized spatial MDX (Multidimensional Expressions) queries. The approach detects implicitly the preferences and needs of SOLAP users using a spatiosemantic similarity measure. The approach is described theoretically and validated by experiments. Background
Data warehouses (DW) are being considered as efficient components of decision support systems [7]. They are usually structured according to the multidimensional structure (also called a cube),
which facilitates a rapid navigation within different levels of data granularity (from coarser to finer level and vice versa). Spatial Data warehouses (SDW) store a large amount of historized spatial data which have specific characteristics such as topology and direction. SDW can be explored by SOLAP systems (Spatial OnLine Analytical Processing) to enable spatial online analysis. SOLAP systems combine both GIS and OLAP (On-Line Analytical Processing) technologies. They offer, to users, opportunities for spatial analysis of geo-localized data by allowing them to visualize and navigate through aggregated spatial data according to a set of dimensions with different levels of granularity. SOLAP users can exploit spatial data warehouses by launching a sequence of MDX (Multidimensional Expressions) queries.
Conclusion
In this paper, we propose a personalization of SOLAP systems through a recommendation approach. The approach assists the user in spatial data warehouse exploitation through the recommendation of personalized MDX queries. The approach (1) detects the preferences and the analysis objectives of the user using a collaborative filtering technique, and (2) applies a spatio-semantic similarity measure between MDX queries to compare the analytical objectives of the users and their preferences. We presented a theoretical framework detailing the various phases of the approach namely (1) log file filtering, (2) generation of candidate queries (3) ranking recommendations, and (4) presentation of recommendations. Each step is explained by detailed algorithms presenting how these phases can be implemented.
