Thread: Logistic Regression versus Bayesian

I have written a bot/spider that collects webpages.

I'm trying to predict whether a webpage is about a conference - a boolean outcome.

I'm wondering what the difference is between using logistic regression (on each keyword) versus using a Bayesian filter. I know Bayesian filters are used to detect spam and are capable of learning. But I'm wondering whether there is a difference in the predictive power of each method if you already had a data set to work with?

Originally Posted by akreider

I have written a bot/spider that collects webpages.

I'm trying to predict whether a webpage is about a conference - a boolean outcome.

I'm wondering what the difference is between using logistic regression (on each keyword) versus using a Bayesian filter. I know Bayesian filters are used to detect spam and are capable of learning. But I'm wondering whether there is a difference in the predictive power of each method if you already had a data set to work with?

In any given situation, the expected predictive performance of these methods is not likely to be exactly equal. Determining which performs better will require actually building both on the data, for comparison.

Getting away from predictive performance, there are other reasons to prefer one method over another. Naive Bayes models, for example, handle missing data with ease.


-Will Dwinnell
Data Mining in MATLAB