Before we understand what are classifiers we first have to understand classification. Classification simply means to classify things/objects into similar groups.

Suppose we have 5 ice-creams and 7 cupcakes with us. How do we know their exact number, the simple answer is by counting them (basic Maths), which our parents and teachers have taught us long back. This distinction is easy for us and even for the kids but it is not easy for the machines!

For Machines to know ice creams are different from cupcakes they need to be trained with help of a lot of examples, which can be called input to the machine. The machine will then learn to classify the classes into separate groups.

This is done with the help of Classifiers. Classifiers are what we call models. It is called a classifier by Scikit learn library of Python. The classifiers look for patterns in the data.

The above example is a Binary Classification Problem in Machine Learning. I can call it Supervised Machine Learning because the output is already known. The classifiers have to simply learn to match the input with the right output. For this we have Classifiers.

Classes are also known as labels/target variables in Machine Learning. we train the machines to classify ice-creams and
cupcakes correctly, but this holds a lot of challenges for the machine, and we humans have to solve them which we shall
discuss later.

For classification problems in Machine Learning, we have classifiers. We can have binary (two classes) as well as multi-label
(more than two classes) classification.
The classifiers help to understand the class and help to classify the data we give as input to the machines.
The classifiers need many inputs as examples to sufficiently get trained; after the training, the classifier is tested
on unknown examples.

The accuracy and effectiveness of the classifier is determined by how correctly the classes are
classified.

Types of Classifiers:
1) Linear classifiers
2) Non-Linear Classifiers

NOTE: To Understand better you can watch the turorial https://youtu.be/wucKtqkKciw

Linear Classifiers and Non-Linear Classifiers

The classes that are linearly separable uses linear classifiers like Linear Regression, Support Vector Machine (with Linear kernel)Naive Bayes, Logistic Regression.

When a given problem cannot be solved linearly, or by separating them we use non-linear classifiers.
Quadratic Discriminant Classifier, Multi-layer Perceptron (most deep networks), Decision Trees, Random Forest, K-Nearest Neighbour.

Which is the best Machine Learning Classifier

There is no best classifier but we have to check for multiple classifiers and compare them, choose 2 or 3 models which
give the best result. The classifier giving the best result is chosen for making predictions on the test data.

Importance of classifiers in Machine Learning

1)Help to build the model.
2)Understand Patterns in the data.
3)Helps to map the input with the output in the case of supervised machine learning.
4)They are useful for solving classification problems like spam detection, credit card default, detecting cancer cells,
classifying wines, classifying flower species, animal species, etc.