Before going forward let me tell you, I am creating a Machine Learning Series, where I write What I have learned so far. So, Please do take benefit of it and please let me know your reviews because that's very important for me to improve.

Various types of Machine Learning Systems are used to solve various types of problems. Here, We are going to understand how to choose a perfect Machine Learning System to solve your problem.

Machine Learning Systems are categorized based on the following things:

• Human Supervision
• Incrementally Learning
• Predicting or Detecting

Now, Let's understand each category.

Human Supervision Based Machine Learning

Here, the meaning of Human Supervision means the data passed to the Machine Learning algorithm is labeled.

What is labeled data?

Your machine learning algorithms are based on the Data that you pass to it for learning. There are two kinds of Data Labeled and not labeled.

If the data you passed has multiple features and their desired solutions then that is Labeled Data and those solutions are labels.

Now Let's see the Machine Learning Systems based on Human Supervision

Supervised Learning

This is a Machine Learning system where we use Labeled data that has features with their desired solutions.

The Supervised Learning Systems solves problems of Classification of data in particular labels and Predicting some values based on the features.

Let me give an example for each of the different problems.

Classification problems are like categorizing emails as Spam or non-spam based on their content. Here the features are the content of the email which is used to classify the emails and spam and non-spam which are labels.

Prediction problems are like predicting the car price based on its different features like mileage, speed, brand, age, etc.

Here are some important Supervised Learning Algorithms :

• Linear Regression
• Logistic Regression
• k- Nearest Neighbor
• Support Vector Machine
• Decision Trees

Unsupervised Learning

In Unsupervised Learning the Data is unlabeled. The data has only the features and using some unsupervised learning algorithms can help you to extract insights from that raw data.

Following are the tasks that can be done using Unsupervised Machine Learning Algorithms:

Clustering

Clustering is a process of finding clusters or separate groups in the Data. Suppose some e-commerce company wants to apply a clustering algorithm to the data of its customer. They may want to create different clusters of their users based on different features from data. By doing this they can target specific clusters to create offers according to their needs.

Some Clustering algorithms are as follows:

• K - Means
• DBSCAN
• Hierarchical Cluster Analysis

Anomaly Detection

This is also one of the important use cases of Unsupervised Machine Learning. Here we detect unusual data and try to study the problematic events. Suppose you have data of transactions of a particular user of a bank. After passing the data to the Anomaly Detection algorithm it detects some unusual behavior in the account. So by detecting the unusual behavior at right time it can prevent frauds to happen.

Algorithms for Anomaly Detection are:

• One-class SVM
• Isolation Forest

Semi-supervised Learning

This is basically the combination of both Supervised and Unsupervised Learning. When we have data that has Labelled as well as unlabeled data. Most of the time unlabeled data is more than labeled data.

Consider photos application in your mobile where it tries to make a group of photos of some people. What it doing is that it is detecting a pattern in unlabeled photos(data) this is an Unsupervised side of the application but then it asks you to label some of the groups and when you label some photos then it tries to detect the label of the photos this is Supervised part of the problem.

Reinforcement Learning

Reinforcement Learning is totally different from whatever you have seen till now. The Learning model which is called an agent will observe the surroundings and performs some action. After acting it will get a reward or penalty( negative-reward). This model learns from these rewards and penalties. Because for the next set of actions it tries to avoid penalties.

The basic overview of the steps that it follows to learn :

Conslusion

Here we learned how machine learning systems are differentiated and seen Machine Learning based on Human Supervision.

So in the next blog, we will study some more Machine Learning Systems which will be based on the Learning capabilities of the model( Read it here ).