As many analysts are aware, R is a powerful language for statistical analysis and visualization. This open source language has top-notch graphics capabilities while offering the flexibility for users to quickly develop custom analyses and to modify them. It also incorporates standard statistical tests, models, and analyses.
These capabilities serve financial services companies well, particularly those with data-intensive research and trading operations that make use of Monte Carlo simulations and other types of computational algorithms for options pricing, the development of risk management models, and other compute-intensive applications.
Yet despite its strengths, R also has some inherent limitations, including memory management. Many R commands can quickly gobble up all available memory. R also has a steep learning curve and isn’t geared for novice users/programmers.
In recognition of these challenges, TIBCO Spotfire has created TIBCO Enterprise Runtime for R (TERR for short), a predictive analytics tool that is native to Spotfire.
TERR provides the R language to financial services companies in a highly scalable and secure outsourced environment. The combination of TERR and Spotfire eliminates the need for any user interface coding. Users can effectively go from install to insight in one day.
The integration of R with Spotfire offers several benefits to analysts, hedge fund managers, and other finance professionals.
Users can create and test models in RStudio and then distribute them directly to research analysts, fund managers, and other financial professionals via Spotfire. This self-service approach bypasses IT bottlenecks, providing investment managers and other decision-makers with direct access to tests and models.
TERR offers investment banks and other financial services companies dramatic productivity and efficiency gains over comparable R-related offerings.
For instance, one client was involved in a benchmark test with five million rows of data. R took 107 seconds to complete the sequence. With TERR, it took just 17 seconds, or less than one-fifth of the required time. Meanwhile, model scoring of 20 million rows for the same client was even more lopsided, with R requiring 84 seconds and TERR just one second to complete.
