StatPatternRecognition on Analysis of HEP and Astrophysics Data
I. Narsky · CERN Document Server (European Organization for Nuclear Research) · 2008
StatPatternRecognition (SPR) is a C++ package for supervised machine learning.Introduced in 2005, it has been used by several HEP and astrophysics collaborations, as well as non-academic researchers, for analysis of complex multivariate data.The package implements powerful classification algorithms such as boosting (three flavors), arc-x4, bagging, random forest, neural networks, decision trees (two flavors), bump hunter (PRIM), multi-class learner, logistic regression, linear and quadratic discriminant analysis, combiner of classifiers, and others.It also offers a suite of tools for data analysis: estimation of variable importance, bootstrap, cross-validation, computation of data moments, multivariate goodness-of-fit estimation, and others.SPR is a standalone package with an optional dependency on Root for data input/output.The user can access major SPR methods from an interactive Root session by loading the SPR shared library.The package is under active development and shows a growing number of users in the HEP community and elsewhere.The latest source release of the package can be obtained under General Public License from Sourceforge [1].A full list of notes and talks about the package can be found on the author's web page [2].