Introduction

In general we can say that machine learning tasks can be broken down into three main areas:

• supervised
• unsupervised
• reinforcement learning.

Supervised learning learns from a set of labelled input data, and based on what was learnt, the system will make classifications of new, previously unseen data, or trying to forecast an outcome. Unsupervised learning brings out the hidden patters in a data set. In other words, the features are not explicitly given, as is the case with supervised learning. This type of learning looks to discover the similarity in the data set for clustering tasks, or to uncover the relationships in the data. Reinforcement learning is a process which learns through trial and error, using feedback to improve the process.

Supervised learning can further be broken down into classification and regression algorithms. The main focus here will be on the classification side.

Classification tasks in the world of ML are very common. Classification tasks require using ML algorithms to assign a certain class label to a given data set, as long as that data set comes from the problem domain. The common example given when one starts to look at classification problems is that of an email spam filter. Using natural language processing (NLP), an ML model can access the likelihood on an incoming email being spam or not.

There are a number of classification algorithms available, both for binary classification and multi-class classification. The algorithms we will look at are:

1. Decision Trees
2. Random Forest
3. Support Vector Machines (SVM)
4. K-Nearest Neighbor (KNN)
5. Logistic Regression

Algorithm Discussion

In this section we will discuss the main algorithms used for classification tasks.

Decision Tree

A decision tree will build tree branches in a hierarchical fashion. To understand what is going on within the algorithm, each branch can be considered an if-else statement.
https://towardsdatascience.com/decision-trees-in-machine-learning-641b9c4e8052

Sources