Читать книгу Fundamentals and Methods of Machine and Deep Learning - Pradeep Singh - Страница 19

Logistic regression is well-known ML algorithms, which is under the SML technique. It is utilized for anticipating the dependent factor by making use of a given set independent factor, it is used for the classification problems, and it is dependent on the idea of probability. Logistic regression calculates the yield of a dependent variable. Thus, the outcome is a discrete value. It may be either yes or no, zero or one, and valid or invalid [3, 7]. However, instead of giving the definite value as 0 and 1, it provides the probabilistic values which lie in the range of 0 and 1. For instance, consider that you are being given a wide scope of riddles/tests trying to comprehend which concept you are acceptable at. The result of this investigation would be considered a geometry-based issue that is 70% prone to unravel. Next is the history quiz, the chance of finding a solution is just 30%. Consider an event of detecting the spam email. LR is utilized for this event; there is a constraint of setting a limit depending on which classification is possible. Stating if the class is spam, predicted consistently is 0.4 and the limit is 0.5, the information is categorized as not a spam mail, which can prompt the outcome progressively. Logistic regression is classified as binary, multinomial, and ordinal binary can have only two possible values either yes or no or true or false where multinomial can have three or more possible values and Ordinal it manages target factors with classifications. For instance, a grade can be arranged as “very poor”, “poor”, “great”, and “excellent”.

Logistic regression is well defined as [16].

(1.5)

Figure 1.3 Logistic regression [3].

Figure 1.3 shows the function curve between the values 0 and 1.