Repository files navigation

In this work, we tested the capabilities of various Bayesian networks structures (mainly Naive Bayes and augmented Naive Bayes) in a classification task, over the standard Adult dataset, which aims at separating people whose income is greater than 50 thousands dollars per year from the rest.

In order to play with the provided Jupyter notebook and test the various classifiers, it is necessary to follow these steps:

• InstallPython 3.8 on your system
• Optionally create a virtual environment in the root directory of the project (python3 -m venv venv) and activate it (source venv/bin/activate)
• Install the required dependencies (pip install -r requirements.txt)

• Naive Bayes (NB): Implementation given bypgmpy
• Tree-Augmented Naive Bayes (TAN): Implementation taken by a pending pull request on thepgmpy repository
• BN-Augmented Naive Bayes (BAN): Custom implementation (slow and buggy)
• Forest-Augmented Naive Bayes (FAN): Custom implementation

The Adult dataset was downloaded from theUCI Machine Learning Repository and placed inside thedataset folder.

The whole project was written in the Jupyter notebookclassify.ipynb, while the custom structural learning algorithms are implemented in theestimators.py file.

Moreover, a complete overview of the whole data pre-processing, classification and evaluation pipeline can be found in thereport.pdf file, inside thereport folder.