Automated comparison of scikit-learn binary classifiers.
GridSearchClassifier is a tool to find the best performing classifier (according to accuracy/F1, train time and test/deployment time) using a given labelled training set and labelled test set. The user can input a list of scikit-learn classifiers and classifiers compatible with scikit-learn. The tool searches through classifier provided in the list. The tool can be used both to choose between different algorithms and to choose hyperparameters for a specific algorithm.
- On smaller data frames.
- Initial exlploratory of different algorithm performance.
- To rule out algorithms.
- Hyperparameter selection.
There are examples of using GridSearchClassifier on face recognition data in /example_binary_face_recognition.py