ASReview Ensemble Classifiers Extension

This extension adds a new set of classifiers by ensembling different basic classifiers such as Naive Bayes (NB), Logistic Regression (LR) and Random Forest (RF).

Getting started

To install this extension, clone the repository to your system and then run the following command from inside the repository.

pip install .

or you can directly install it from GitHub using

pip install git+https://github.com/rohitgarud/Asreview-Ensemble-Classifiers.git

Usage

Four different ensemble classifiers are currently available: ensemble_nb_lr (NB+LR), ensemble_nb_rf (NB+RF), ensemble_lr_rf (LR+RF), ensemble_nb_lr_rf (NB+LR+RF). Simulations can be performed using the simulation mode from ASReview CLI using:

asreview simulate example_data_file.csv -m ensemble_nb_lr -e tfidf

If Naive Bayes is part of the ensemble, we have to use the TFIDF features. However, if only the Logistic and Random Forrest ensemble is used, we can use other features such as Doc2Vec or SBERT.

Also, a comprehensive simulation study can be performed using the ASReview Makita Extension (follow the instructions on the extension GitHub page). One example of simulation using is a comparison of NB and Ensemble of NB and LR classifiers can be performed using:

asreview makita template multiple_models --classifiers nb ensemble_nb_lr --feature_extractors tfidf -f jobs.bat

Four different ensemble strategies are available mean, max, multiply and random. The default is the multiply ensemble strategy. To use other settings, you have to use the Python API

License

Apache 2.0 license