You can use it to classify company e-mails, customer support transcripts, AP wire articles, the police blotter, medical documents, movie reviews, whatever!

This issue, handling categorical data, that is, data that's grouped by a finite number of labels without inherent numeric equivalents, is one that constantly nips at data miners' heels.

You have to be smart about pulling relevant predictors out of the dataset.

mean big business, more complex measures of prestige, infl uence, and centrality have evolved to account for such bad behavior.

These predictors are often called model features or independent variables, while the thing we're trying to predict “Pregnant (yes/no)?” would be the dependent variable in the sense that its value is dependent on the independent variable data we're pushing into the model.

business receives feedback on how their audience is engaging at the individual level through click tracking, online purchases, social sharing, and so on.

But where forecasting and supervised machine learning differ greatly is in their canonical problem spaces.

At a simplistic level, you feed a supervised AI algorithm some historical data, purchases at Target for example, and you tell the algorithm, “Hey, these purchases were from pregnant people, and these other purchases were from not-so-pregnant people.”

It is easy, especially with sparse datasets (only a few observations), to get a model that fits quite well but whose fit is statistically insignificant, meaning that the relationship between the features and the independent variable may not actually be real.