Machine Learning: 2 Books in 1: Machine Learning for Beginners, Machine Learning Mathematics. An Introduction Guide to Understand Data Science Through the Business Application

Generating predictions using logistic regression
In here, you can simply plug in the measurements into the "logistic
regression" equation and calculate the outcome to generate predictions with
the "logistic regression" model. Let's take a look at an example to solidify
this concept. Let's assume there is a model that is capable of generating
predictions if an individual is masculine or woman depending on with
fictitious values of their height. If the value of the height for an individual is
set as 150 cm, would the individual be predicted as a male or female?
Assuming we have already discovered the values of coefficients "b0=
-100" and "b1= 0.6". By leveraging the above equation, the probability of
male with a height of 150 cm or "P(male|height=150)" can be easily
calculated. The function EXP() will be used for "e" because if you log this
instance into your spreadsheet, this is what you can use:
“y = e^(b0 + b1*X) / (1 + e^(b0 + b1*X))”
“y = exp(-100 + 0.6*150) / (1 + EXP(-100 + 0.6*X))”
“y = 0.0000453978687”
Or a near "0" probability male is the gender of that specific person.
In theory, probability can simply be used. But since this is a "classification"
algorithm and we want a sharp outcome, the probabilities can be tagged on
to a binary class value. For instance, the model can predict "0" if "p (male)

---

< 0.51" and predict "1" if "p (male) >= 0.5". Now that you know how
predictions can be generated using "logistic regression", you can easily pre-
process the training data set to get the most out of this technique. The
assumptions made of the distribution and relations within the data set by the
"logistic regression" technique are nearly identical to the assumptions made
in the "linear regression" technique.
A lot of research has been done to define these hypotheses and to use
accurate probabilistic and statistical language. It is recommended to use
these as thumb rules or directives and try with various processes for data
preparation.
The ultimate goal in "predictive modeling" machine learning initiatives
is the generation of highly accurate predictions rather than analysis of the
outcomes. Considering everything some assumptions could be broken if the
designed model is stable and has high performance.