DROLAP - A Dense-Region Based Approach to On-Line Analytical Processing

ROLAP Relational OLAP and MOLAP Multidimensional OLAP are two opposing techniques for building On line Analytical Processing OLAP systems MOLAP has good query performance but su ers when the data distribution in the multidimensional data cube is sparse ROLAP can be built on mature RDBMS technology but its performance is not as competitive Many data warehouses contain sparse but clustered multidimensional data We propose a dense region based OLAP DROLAP system which surpasses both ROLAP and MOLAP in space e ciency and query performance DROLAP applies the MOLAP approach on the dense regions discovered in the data and handles the remaining small percentage of sparse points with the ROLAP approach The core of building a DROLAP system lies in the mining of dense regions in a data cube We have de ned the dense region mining problem as an optimization problem We show that conventional clustering techniques are not suitable for this problem and have developed an e cient index based algorithm EDEM to compute dense regions Extensive performance studies have been performed Our results clearly show that the DROLAP approach is superior than both the MOLAP and ROLAP approaches The results also show that EDEM is e cient and e ective in locating dense regions