A comparative study of support vector machine and logistic regression article · January 021 citations reads 11 authors

1.3.2 Logistic Regression 
The logistic regression is a predictive tool that measures the relationship between the categorical 
dependent variable and one or more independent variables by estimating probabilities using a 
logistic function. Logistic regression models the likelihood of two categories, it is used to evaluate 
the likelihood of gender where the response gender falls within a category say male or female. 
Logistic regression can be binomial or multinomial. In the binomial or binary logistic regression, 
the outcome can have only two possible types of values (example, “YES” or “No”, PASS” or 
“FAIL”). Multinomial logistic measures where the result can have three or more possible types of 
values (example, “GOOD”,” BETTER”,” BEST”). Generally, the result is known as “0” and “1” 
in binary logistic regression. Straight line is fitted to our observations in linear regression while 
we fit an S shaped curve called sigmoid to our observations in logistic regression. Sigmoid function 
takes 1 as maximum and 0 as minimum which indicates why Y-axis ranges from 0 to 1. The 
formula for the sigmoid function is stated in (1.1). 
(1.1) 

---

O
RIENTAL JOURNAL OF SCIENCE & ENGINEERING VOL -2, ISS-1, FEB - 2021 
www.ojse.org
ojse©2019 
Page 88 
Figure 2: Graphical representation of logistic regression 
1.3
Statement of Problem 
Classification has been a very important aspect of data analysis and has been applied to several 
fields like medicine, finance, agriculture, science and technology, economics, etc. There are 
different ways of carrying out classifications namely the linear discriminant analysis (LDA), the 
logistic regression (LR) and the support vector machine (SVM), to mention but a few. This 
research concentrated largely on the SVM and LR. These methods have different approaches to 
classifying data, consequently, it is expected that they will perform differently based on how well 
they classify new observations.