U.S. Department of Energy

Pacific Northwest National Laboratory


glmnetLRC is an R package that makes it easy to construct a binary classifier from virtually any number of quantitative predictors that will assign an example, or observation, to one of two classes. It extends the glmnet package by making it possible to train lasso or elastic-net logistic regression classifiers (LRC's) using a customized, discrete loss function to measure the classification error.  This allows users to assign unique  loss values to false positive and false negative errors. The logistic regression parameter estimates are obtained by maximizing the elastic-net penalized likelihood function that contains several tuning parameters. These tuning parameters are estimated by minimizing the expected loss, which is calculated using cross validation. This approach was originally implemented to automate the process of determining the curation quality of mass spectrometry samples.

Additional Staff: 
Alex Venzin
Principle Investigator: 
| Pacific Northwest National Laboratory