Logistic regression was developed during the 19th century to study the growth of population and some specific types of chemical reactions, and the first person to formally define it was the Belgian statistician Pierre François Verhulst, who published in 1837 four pages about it within his mentor's publication, Correspondance Mathématique et Physique.
Starting from this first publication, a lot of others followed, paired with extensive use of the model in real-life domains far from the original one, such as fraud detection and the estimation of the probability of default.
The intuition behind logistic regression starts exactly where linear regression stops—solving the problem of estimated values outside the natural domain of our response variable. It start from the typical problem of having a response variable that can pertain to two alternative categories, either zero or one, and the need for some record to one of those categories...