predict.py is the Python replacement code that does the same job as glmNetPred.R
Used the python package glmnet_py as described on https://github.com/bbalasub1/glmnet_python. It is available only for Ubuntu + Python 3.5 as of now.
This was done as the R code was giving some convergence errors for the same models, and was taking an extremely long time to complete.
FeatureVectorNER.csv is the feature file with NER features. Here are the modifications made to this file:
- 2nd column, i.e. the negated target column is removed MANUALLY.
- Group column is replaced by the NER-features obtained by the Java code. "Individual" column is then removed MANUALLY.
FeatureVectorNotNER.csv is the original features file with Group and Individual features. Here are the modifications made to this file:
- 2nd column, i.e. the negated target column is removed MANUALLY.
The other files in the format train-Strat-NER-{}.csv or test-Strat-NER-{}.csv represent the train and test sets for a particular split for the features with NER.
This is similar in the case for NotNER feature files.