Gaussian Naive Bayes Classifier for Income Data Set

GNB classifier trained to predict individual's income

Data Set

Data set used in this implementation is Adult Data Set of UCI machine learning repository, visit the website for more information.

Implementation details

• Ignore any unknwon attributes that marked as "?" in the dataset and sum up the probability of the occurance of discrete attribute to 1
• Assume the result of log 0 is negative infinite.

Training Result

The prediction error obtained from this implementation is 16.90%, which is larger than the result descripted on the website, which is 16.12%. Probably it is because I didn't use smoothing in this implementation.

Libaries

• Use Numpy for scientific computing.