Welcome to the Iguanas documentation!¶
What is Iguanas?¶
Iguanas is a fast, flexible and modular Python package for:
Generating new fraud-capture rules using a labelled dataset.
Optimising existing rules using a labelled or unlabelled dataset.
Combining rule sets and removing/filtering those which are unnecessary.
Generating rule scores based on their performance.
It aims to help streamline the process for developing a deployment-ready rules-based system (RBS) for binary classification use cases.
What are rules and rule-based systems (RBS)?¶
A rule is a set of conditions which, if met, trigger a certain response. An example of a rule is one that captures a particular type of fraudulent behaviour for an e-commerce company:
If the number of transactions made from a given email address in the past 4 hours is greater than 10, reject the transaction.
An RBS is one that leverages a number of these rules to provide a certain outcome.
For example, an e-commerce company might employ an RBS to accept, reject and review its transactions.
The pros and cons of an RBS¶
As with any approach, there are pros and cons to RBS’s.
Pros
Rules are intuitive, so the outcome given by the RBS is easy to understand.
RBS’s are flexible, since rules can be quickly added to address new behaviour.
Rules can be built using domain knowledge, which may capture behaviour an ML model would have missed.
Cons
Linked to the last pro - domain knowledge is usually required to build rules. If you don’t have the domain knowledge, creating rules can be difficult.
Generating these rules can be challenging and time consuming, especially if a data-guided approach is used.
Difficult to tweak existing rules to address new trends.
The solution – Iguanas!¶
Iguanas addresses the cons of an RBS for binary classification problems:
Iguanas only requires a historic dataset to generate rules – similar to the requirements of an ML model.
Iguanas quickly and easily generates high performance rules and utilizes an API that is familiar to most data scientists - Sklearn’s fit/transform methology.
Iguanas’ rule optimization module allows the user to tweak the thresholds of current rules using a labelled dataset.
Iguanas also has a host of other modules which help to streamline the RBS set up process.
Getting started¶
The Installation section provides instructions on installing Iguanas on your system. Once installed, take a look at the Examples section for how Iguanas can be used. The API section also provides documentation for the classes and methods in each module.