Skip to content

Automatic evaluation of different machine learning models with focus on chemistry-related tasks

License

Notifications You must be signed in to change notification settings

BernalFA/chempred

Repository files navigation

ChemPred

ccds

 

ChemPred is a tool for automatic evaluation of different machine learning models with focus on chemistry-related tasks. It takes advantage of the pipelining capabilities offered by scikit-learn, imbalanced-learn, and scikit-mol. Thus, model performance comparisons on pipelines using several molecular featurization are one of its intended uses.

Important

In its current implementation, ChemPred does not offer hyperparameter tuning. Models and pipelines are created using default values.

Usage

Minimum example (classification mode):

from chempred.experiment import ClassificationExplorer

# set up exploration experiment
experiment = ClassificationExplorer()
# run evaluation (automatic pipeline creation, fitting, and scoring)
experiment.evaluate(smiles_train, smiles_test, y_train, y_test)
# check results (pandas DataFrame)
experiment.results_.head()
# output
> Algorithm                  | Balancing method    | Molecular Transformer           | balanced_accuracy | Time
> LogisticRegression         | RandomUnderSampler  | MACCSKeysFingerprintTransformer | 0.800             | 0.133
> MLPClassifier              | None                | AvalonFingerprintTransformer    | 0.811             | 2.315
> RidgeClassifier            | SMOTE               | AvalonFingerprintTransformer    | 0.766             | 0.282
> GaussianProcessClassifier  | None                | MorganFingerprintTransformer    | 0.768             | 6.099
> RandomForestClassifier     | RandomUnderSampler  | AvalonFingerprintTransformer    | 0.815             | 0.396

Installation

At the moment, only installation from source is possible:

git clone https://github.com/BernalFA/chempred.git
cd chempred
pip install

License

The content of this repo is licensed under the MIT license conditions.

About

Automatic evaluation of different machine learning models with focus on chemistry-related tasks

Resources

License

Stars

Watchers

Forks

Packages

No packages published