Logistic regression is one of the simplest ways to build classifiers: models that predict binary outcomes (in contrast to regression models, which predict numerical values).
There are many ways to ask the question “why logistic regression?”. One natural question to ask is: if we already know how to build linear regression models, but now are trying to build classifiers, why not simply use a linear regression model and round the prediction to a binary value?
TBF: demo of why this idea doesn’t work.
TBF. <div id="div-plot2"></div>