Smart-Choice: Decision-Making Analysis Toolkit

Author

Prof. Juan David Velásquez-Henao, MSc, PhD
Universidad Nacional de Colombia, Sede Medellín.
jdvelasq@unal.edu.co

What is it?

Smart-Choice is a Python package for Decision-Making Analysis using decision trees. Smart-Choice allows the user to define a decision tree directly in Python. The best experience can be obtained when the package is used in a notebook inside of Jupyter Lab or Google Colab. Smart-Choice has no limits in the size of the decision tree created, and it can efficiently run large trees. Different reports are available to facilitate tree analysis.

Main Features

The package allows the user to define the following types of nodes in a decision tree:

  • Chance nodes.

  • Decision nodes.

  • End or Terminal nodes.

In the package, all model values and probabilities are entered directly as node properties using typical data structures in Python. Thus, an user with a basic knowledge of the programming language can use effectively the package.

A run of the decision tree can be used using monetary expected values, but, the following utility functions can be used to represent risk adversion:

  • Exponential.

  • Logarithmic.

Different types of analysis can be conducted easily, including:

  • Decision analysis.

  • Sensitivity analysis.

  • Risk analysis.

For the terminal of end nodes, the user must supply Python functions to evaluate the value of the node. This feature allows the user to use all capacity of Python programming language. It is possibe to write functions to run a complete Monte Carlo simulation using other packages as scipy. In other scenarios, it is possible to build complex predictive models that feed the decision model using, for example, scikit-learn. Other great adventage of the Smart-Choice is velocity where it is compared with spreadsheets; it is possible to run complex models in a fraction of the time required when a spreadsheet is used.

Release Information