erexplain 0.1.2

library to explain models for Entity Resolution

ExplainER

ERExplain is a package to explain any Entity Resolution model, using a Black-Box approach

Key Features

The explanations can be provided into two different ways:

• Attribute-Level: the system shows to the user a histogram reporting the importance of the attributes with respect to the prediction. The importance is a number in [0,1]
• Value-level: the system shows to the user, for each chosen attribute set/single attribute A, the most influential values for A, with respect to the prediction

Tech

Dillinger uses some well-know libraries:

• Pandas: To create perturbation
• Numpy
• MLXtend: To create explanations at value level

Installation

To install the package simply run

$ pip install erexplain