Building Standards Based Analytics for Financial Services

The Grid has developed standards based infrastructures andservices and platforms relevant to data mining. For example, the EuropeanData Grid (EDG) provides a Relational Grid Monitoring Architecture (R-GMA) for distributed resources that expose a relational model with SQLsupport to provide static as well as dynamic information about Gridresources and for use within application monitoring. The relational model R-GMA is very flexible and allows complex queries that make use ofinformation in multiple objects. R-GMA makes information from Producersavailable to Consumers as relations (tables). This paper describes theimplementation of a ScanOnce algorithm in SQL for quick association rulemining and the development of a data mining infrastructure JetGrid. Thearchitecture of JetGrid is designed to be compatible with lower-level gridmechanisms since it is to operate on top of R-GMA. JetGrid for quickknowledge discovery was preliminarily prototyped and it is extensible, usingan object-oriented design that was coded in C++. Mining agents will bestaged to one or more computing elements on the EDG. The agent alsoperforms acquisition of data. Intensive tests are being carried out usingGridFTP and other Globus transport mechanism for shipping raw data Dacross the grid to one node for processing (MD: Move Data Model) andprocessing the data locally until a result R is obtained and ship the result toone node for further processing (MR: Move Results Model), which providessome indication of the performance of the JetGrid data mining infrastructureon top of EDG.