instancelib 0.5.2

A generic interface for datasets and Machine Learning models

A generic interface for datasets and Machine Learning models

---

instancelib provides a generic architecture for datasets and machine learning algorithms such as classification algorithms.

© Michiel Bron, 2021

Quick tour

Load dataset: Load the dataset in an environment

import instancelib as il
text_env = il.read_excel_dataset("./datasets/testdataset.xlsx",
                                  data_cols=["fulltext"],
                                  label_cols=["label"])

ds = text_env.dataset # A `dict-like` interface for instances
labels = text_env.labels # An object that stores all labels
labelset = labels.labelset # All labels that can be given to instances

ins = ds[20] # Get instance with identifier key  `20`
ins_data = ins.data # Get the raw data for instance 20
ins_vector = ins.vector # Get the vector representation for 20 if any

ins_labels = labels.get_labels(ins)

Dataset manipulation: Divide the dataset in a train and test set

train, test = text_env.train_test_split(ds, train_size=0.70)

print(20 in train) # May be true or false, because of random sampling

Train a model:

from sklearn.pipeline import Pipeline 
from sklearn.naive_bayes import MultinomialNB 
from sklearn.feature_extraction.text import TfidfTransformer, CountVectorizer

pipeline = Pipeline([
     ('vect', CountVectorizer()),
     ('tfidf', TfidfTransformer()),
     ('clf', MultinomialNB()),
     ])

model = il.SkLearnDataClassifier.build(pipeline, text_env)
model.fit_provider(train, labels)
predictions = model.predict(test)

Installation

See installation.md for an extended installation guide.

Method	Instructions
pip	Install from PyPI via pip install instancelib.
Local	Clone this repository and install via pip install -e . or locally run python setup.py install.

Documentation

Full documentation of the latest version is provided at https://instancelib.readthedocs.org.

Example usage

See usage.py to see an example of how the package can be used.

Releases

instancelib is officially released through PyPI.

See CHANGELOG.md for a full overview of the changes for each version.

Citation

@misc{instancelib,
  title = {Python package instancelib},
  author = {Michiel Bron},
  howpublished = {\url{https://github.com/mpbron/instancelib}},
  year = {2021}
}

Library usage

This library is used in the following projects:

• python-allib. A typed Active Learning framework for Python for both Classification and Technology-Assisted Review systems.
• text_explainability. A generic explainability architecture for explaining text machine learning models
• text_sensitivity. Sensitivity testing (fairness & robustness) for text machine learning models.

Maintenance

Contributors

• Michiel Bron (@mpbron)