Python Record Linkage Toolkit

The Python Record Linkage Toolkit is a library to link records in or between data sources. The toolkit provides most of the tools needed for record linkage and deduplication. The package contains indexing methods, functions to compare records and classifiers. The package is developed for research and the linking of small or medium sized files.

This project is inspired by the Freely Extensible Biomedical Record Linkage (FEBRL) project, which is a great project. In contrast with FEBRL, the recordlinkage project uses pandas and numpy for data handling and computations. The use of pandas, a flexible and powerful data analysis and manipulation library for Python, makes the record linkage process much easier and faster. The extensive pandas library can be used to integrate your record linkage directly into existing data manipulation projects.

One of the aims of this project is to make an easily extensible record linkage framework. It is easy to include your own indexing algorithms, comparison/similarity measures and classifiers.

Basic linking example

Import the recordlinkage module with all important tools for record linkage and import the data manipulation framework pandas.

import recordlinkage
import pandas

Load your data into pandas DataFrames.

df_a = pandas.DataFrame(YOUR_FIRST_DATASET)
df_b = pandas.DataFrame(YOUR_SECOND_DATASET)

Comparing all record can be computationally intensive. Therefore, we make set of candidate links with one of the built-in indexing techniques like blocking. In this example, only pairs of records that agree on the surname are returned.

indexer = recordlinkage.Index()
indexer.block('surname')
candidate_links = indexer.index(df_a, df_b)

For each candidate link, compare the records with one of the comparison or similarity algorithms in the Compare class.

c = recordlinkage.Compare()

c.string('name_a', 'name_b', method='jarowinkler', threshold=0.85)
c.exact('sex', 'gender')
c.date('dob', 'date_of_birth')
c.string('str_name', 'streetname', method='damerau_levenshtein', threshold=0.7)
c.exact('place', 'placename')
c.numeric('income', 'income', method='gauss', offset=3, scale=3, missing_value=0.5)

# The comparison vectors
feature_vectors = c.compute(candidate_links, df_a, df_b)

Classify the candidate links into matching or distinct pairs based on their comparison result with one of the classification algorithms. The following code classifies candidate pairs with a Logistic Regression classifier. This (supervised machine learning) algorithm requires training data.

logrg = recordlinkage.LogisticRegressionClassifier()
logrg.fit(TRAINING_COMPARISON_VECTORS, TRAINING_PAIRS)

logrg.predict(feature_vectors)

The following code shows the classification of candidate pairs with the Expectation-Conditional Maximisation (ECM) algorithm. This variant of the Expectation-Maximisation algorithm doesn't require training data (unsupervised machine learning).

ecm = recordlinkage.ECMClassifier()
ecm.fit_predict(feature_vectors)

Main Features

The main features of the Python Record Linkage Toolkit are:

• Clean and standardise data with easy to use tools
• Make pairs of records with smart indexing methods such as blocking and sorted neighbourhood indexing
• Compare records with a large number of comparison and similarity measures for different types of variables such as strings, numbers and dates.
• Several classifications algorithms, both supervised and unsupervised algorithms.
• Common record linkage evaluation tools
• Several built-in datasets.

Documentation

The most recent documentation and API reference can be found at recordlinkage.readthedocs.org. The documentation provides some basic usage examples like deduplication and linking census data. More examples are coming soon. If you do have interesting examples to share, let us know.

Installation

The Python Record linkage Toolkit requires Python 3.5 or higher (since version >= 0.14). Install the package easily with pip

pip install recordlinkage

Python 2.7 users can use version <= 0.13, but it is advised to use Python >= 3.5.

The toolkit depends on popular packages like Pandas, Numpy, Scipy and, Scikit-learn. A complete list of dependencies can be found in the installation manual as well as recommended and optional dependencies.

License

The license for this record linkage tool is BSD-3-Clause.

Citation

Please cite this package when being used in an academic context. Ensure that the DOI and version match the installed version. Citatation styles can be found on the publishers website 10.5281/zenodo.3559042.

@software{de_bruin_j_2019_3559043,
  author       = {De Bruin, J},
  title        = {{Python Record Linkage Toolkit: A toolkit for
                   record linkage and duplicate detection in Python}},
  month        = dec,
  year         = 2019,
  publisher    = {Zenodo},
  version      = {v0.14},
  doi          = {10.5281/zenodo.3559043},
  url          = {https://doi.org/10.5281/zenodo.3559043}
}

Need help?

Stuck on your record linkage code or problem? Any other questions? Don't hestitate to send me an email ([email protected]).