How reduce the View Selection Problem through the CoDe Modeling

This is the authors' abstract. We don't add key points for this paper.

Big Data visualization is not an easy task due to the sheer amount of information contained in data warehouses. Then the accuracy on data relationships in a representation becomes one of the most crucial aspects to perform business knowledge discovery. A tool that allows to model and visualize information relationships between data is CoDe, which by processing several queries on a data-mart, generates a visualization of such data. However on a large data warehouse, the computation of these queries increases the response time by the query complexity. A common approach to speed up data warehousing is precompute a set of materialized views, store in the warehouse and use them to compute the workload queries. The goal and the objectives of this paper are to present a new process exploiting the CoDe modeling through determining the minimal number of required OLAP queries and to mitigate the problem of view selection, i.e., select the optimal set of materialized views. In particular, the proposed process determines the minimal number of required OLAP queries, creates an ad hoc lattice structure to represent them, and selects on such structure the views to be materialized taking into account an heuristic based on the processing time cost and the view storage space. The results of an experiment on a real data warehouse show an improvement in the range of 36-98% with respect the approach that does not consider materialized views, and 7% wrt. an approach that exploits them. Moreover, we have shown how the results are affected by the lattice structure.