# Unsupervised Machine learning

Unsupervised Machine learning is a learning in which machine/model learn a set of data without any supervision or any pre-training where data is not labeled and classified. In this type of learning model act on the data without any guidance.

Here mode understand the unsorted information and group them according to the matching pattern, similarities, color, and size,

Unlike supervised learning, no teacher teach that means no training will be given to the model. Therefore, model himself discover the hidden pattern in unlabelled data without any guidance.

In our life, we have seen so many cases whereas a human we apply Unsupervised learning.

Instance 1 suppose we have some group of Pets lets say dog and cat and you don’t know what they are and what there name (no prior knowledge),  then how we categorize them so we group them based on their color, Hight and look we can easily make two subgroup one have all dog and one group having all cat.

Instance 2 Suppose you have one fruits basket which having different fruits. your task is to group them without even any pre-knowledge of any fruit name infect you have seen the first time so how will you organize them what you will do first ?? You will take a fruit and you will organize by considering the physical characteristics of that fruit let say you consider color. Then you will organize all fruit based on their color.

The red color group having apple and cherry. And the green color group having greps and bananas.

in the next step let say you consider physical size as a character. Now you subgroup in red color with big size is apple fruits and red color with small size cherry fruits.

Same rule you will apply to the green color group, green color with big size is banana and green color with small size grapes.

Instance 3 Let’s suppose you are first time watching Cricket match, and you have no  idea about this game, but to understand game you can group players based on different parameter: Players wearing same color cloth are in one team, further Players of one style can be one team (bowler, batsmen, fielders), or on the basis of playing style (Right Hand vs Left Hand) or whatever way you would observe can group in subgroups.

Instance 4 Suppose one of your old friends invite you on his family party, where you don’t know anyone and all faces are completely new for you, So as a human how you group them you must start grouping them by there dressing, age group, gender or dressing style etc. In this case without prior knowledge you group which called unsupervised learning in Machine learning terms.

So by that Instance, we have seen without having any pre-knowledge or any label or any training, we have group them correctly. The same approach will be applicable to an unsupervised learning algorithm to categorize the data.

## Unsupervised Machine learning divided into two types of algorithms:

1.Association: Association rule is a method to uncover how items are linked to each other. An association rule is applicable where you want to determine rules that define large portions of your data, such as people that buy A also interested to buy B.

For example, data of a grocery store for common patterns and association rules. For example, the rule could be {bread, milk, water} -> {eggs}, this rule would tell you that if someone bought milk and bread together, then they are also like to buy eggs.

Several purchase forms can be observed. For example:

• The most general deal was of pip and tropical fruits.
• Another general deal was of onions and other vegetables.
• If someone buys a beer can also be instated to purchase beer mug.
• If someone buys meat meals, he is likely to have purchase yogurt as well.
• If someone buys tea, he is likely to have sugar cube fruit as well.

2.Clustering: A clustering problem is where you need to discover the essential groupings in the dataset, such as alliance customers by purchasing behavior.

Clustering can be the most significant unsupervised learning problem; so, as every other problem of this kind, its agreements with finding a similar structure in a collection of the unlabeled dataset.

A layman definition of clustering could be “the method of shaping objects into groups whose members are similar in some way”. hence a cluster is the collection of items which are “similar” between them and are “dissimilar” to the items belonging to other clusters.

### Applications of Clustering

Clustering has a great no. of applications across various domains. Some of the most common applications of clustering are:

1. Medical imaging
2. Recommendation engines
3. Search result grouping
4. Anomaly detection
5. Market segmentation
6. Social network analysis
7. Image segmentation

#### Some real-time applications of Unsupervised Machine learning are as shown below:

1. Social Network Analysis to define groups of friends.
2. Planck Quantum Spectrum
3. Market Segmentation of businesses by industry, location, vertical.
4. Nano Camera Fabrication Technology
5. Generalized Shannon Info Theory to Brain Info Theory

Issues with Unsupervised Machine learning:

1. Unsupervised Learning is difficult as compared to Supervised Learning.
2. Resulted data might not be meaningful because no answer ladled are associated.
3. Result need external expert evaluation.
4. Define an objective role in clustering (internal assessment).

Why Is Unsupervised Machine learning requiring regardless of these issues?

1. Annotating huge datasets is very expensive and manually we can label only a few examples. Example: Speech Recognition
2. There may be a situation where we don’t know how many/what group the data is divided into. Example: Data Mining
3. We may need to use clustering to gain some awareness into the structure of the data before planning a classifier.

### Unsupervised Machine learning Commonly use Algorithms

• k-means clustering, Association Rules

Wrapping up Unsupervised Learning: You have parameters like type, size, the color of something and you code an algorithm to predict that whether it is a plant, fruit, bird, animal or whatever it is. It gives you output by taking some inputs.

Unsupervised Machine learning is the training of the machine using data that is neither labeled nor classified and allowing the model/algorithm to act on that info without guidance. Here the task of the machine is to group unsorted data according to patterns similarities, and differences without any prior training of data.

### 2 thoughts on “Unsupervised Machine learning”

1. Very well written post. I like and and can say this is much better then mine.

1. Thank you Sir, But you are all time Awesome and I inspired from you and your knowledge.