Risk Library is a software package specifically developed for implementation of risk models in insurance and banking.
By using Risk Library modelers can implement their custom capital models very efficiently in a platform of their choice. Currently the supported platforms include R, S-Plus, Matlab, Mathematica, WhiteLight, and SAS9.
Risk Library is written in C++ programming language, and has been subject to comprehensive testing and performance optimization over the years. In addition to extensive unit tests performed by EVMTech, independent testing has been done by validation teams of our clients. The performance is optimized to significantly reduce the computational time that is required to run capital models. This enables the modelers to conveniently test the developed models with different parameter or distributional assumptions.
The main package in the Risk Library are:
Distribution
Distribution Package consist of a comprehensive set of parametric and non parametric distributions such as Lognormal, Pareto, Burr, Weibull, Generalized Pareto, piece-wise linear and constant, or frequency distributions. The package includes functions to compute all relevant statistics of distributions.
Operation
Operations Package consists of distribution operations such as aggregation, sum, mixing, shifting, or scaling, which are implemented using numeric computational techniques.
Fitting
Fitting Package allows fitting loss data to parametric distributions in the distribution package. Fitting supports fitting to losses or quantile values, and truncated fitting. Several goodness of fit tests including KS, Anderson Darling, and log-likelihood function are provided.
Copula
Copula Package consists of several copula models and the Monte Carlo routines to generate joint distribution for specified marginal distributions and copula functions. The supported copulas are Gaussian, Student t, Gumbel, Clayton, Monotonic.
Insurance
The Insurance Package allows applying common (re)insurance contracts to compute the transferred or retained risk distributions. Retained distributions correspond to net risk.
Monte Carlo
The Monte Carlo Package allows Monte Carlo simulations to be performed on any user-defined function of random variables and arguments.
.NET/DLL Link
.NET/DLL Link Package provides the possibility of using AMA Library via .NET and as a dynamic linked library. In particular the library can be linked to many standard platforms such as Mathematica, VB.NET, R Project, MATLAB, or SAS9.
Risk Library is used in conjunction with a modeling platform such as MATLAB. The workflow can be described as follows:
- The modeling platform imports the relevant risk data from the data repository
- Risk Library is linked to the platform
- Risk models are developed using the generic functionality of the platform and the specific functionality of Risk Library
- The computational results are either passed to the data repository or used directly for reporting.
The described operating model provides a simple, flexible, and efficient approach to develop and deploy custom risk models. In addition, the optimized performance of the libraries significantly reduces the computational time, allowing for performance of extensive model analysis and testing.
Typical application areas are operational risk modeling and insurance risk modeling.
Risk Library is developed in C++ and on client request available also in C# and Java. The .NET/DLL Link Package enables Risk Library to be used via .NET and as a dynamic linked library. In particular the library can be linked to many standard platforms such as Mathematica, VB.NET, R Project, MATLAB, or SAS9.