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

Kernel Support vector
Although we will touch on kernel support vectors, later on, they are used to
sort classes that can't be separated with a linear separation line. The
separation line can take on many shapes (linear, nonlinear, polynomial,
radial basis, sigmoid).
Recall the classification by a linear boundary that we just talked about,
where the classification looks something like the following image:

---

In this image, our data can be classified by a straight line that separates the
two distinct categories of data. It would be convenient if data was always
separable like this, but unfortunately, it's not always so neat and tidy in fact,
most of the time you will have to separate the data in a way that looks more
like this:
In this example, the data can’t be separated by a linear boundary line. So
instead we must use a technique called the kernel trick. It uses a measure of
similarity between data points to classify them.
Naïve Bayes
Recall Bayes theorem from the first section on supervised learning. With
Naïve Bayes models, predictors are assumed to be independent. This model

---

is easy to use and helpful in large datasets. Its often employed to help sort
spam emails.
We use Baye’s rule here. The idea of Bayes Rule is that by adding new,
relevant information to what we already know, we can update our
knowledge based on that new information. If we wanted to know the
probability of there being rain this afternoon, we could just figure out what
percentage of days it rains per year. But then we found out that it rained this
morning. How do you think this will affect the probability that it will rain
this afternoon?
So, our ability to predict the probability of something will improve when
we receive more information about the event.
Mathematically, Bayes theorem is expressed as follows:
So, we can classify new data points using Bayes theorem. The way that
works is when we are given a new data point, we calculate the probability
of that data point falling into a category based on the features of that data
point.

---