Linear regression Linear regression is a model of the dependence of the variable
X on one or more other
variables (factors, regressors, independent variables) with a linear dependence function. Linear
regression refers to the task of determining the "line of best fit" through a set of data points
(Figure 1) and has become a simple precursor to non-linear methods that are used to train other
sophisticated methods. Linear regression is the starting point for statistical learning methods,
from it starts the acquaintance with the theme. Because of simplicity and low flexibility linear
regression usually uses in combination with other techniques. For instance of using it in practical
– article from Belarusian researches about measuring blood pressure parameters to determine the
risk of secondary hypertension[2].
Figure 1
Logistic regression Logistic regression predicts the probability of an event from the values of inputs. To do
this the dependent variable
y , which takes only one of two values - as a rule, these are the
numbers 0 (the event did not happen) and 1 (the event happened), and many independent
variables (also called signs, predictors or regressors) - real
X1, X2, ..., Xn , based on the values of
which are required calculate the probability of the adoption of one or another value of the
dependent variable. Logistic regression is a method for constructing a linear classifier, which
allows to evaluate the posterior probabilities of objects belonging to classes. The main idea of
logistic regression is that the space of initial values can be divided by a linear boundary into two
regions corresponding to the classes. If the source data points satisfy this requirement, then they
can be called linearly separable (Figure 2). Logistic regression is often used for predicting
diagnoses in classification manner.
