Machine Learning : How Machine Learning Works

Welcome to the article on how machine learning works.

In the previous article we learnt about:

1. Machine Learning
2. Its applications in the day-to-day life

In this article we will learn more about how machine learning work.

Learning Objectives:

At the end of this topic, you will be able to:

1. Explain what data is and its role in machine learning
2. Explain what is a machine learning model and how it interacts with data to make predictions and recommendations.

Previously we learnt that machine learning is the ability of a machine that is a computer program to learn from past experiences to make predictions or recommendations. The past experiences in the context of a computer program is nothing but the data it has of past events observations or behavior.

So what is data? In simple words any information about anything can be termed as data.

For example 1 :- The features of your smartphone such as screen size, memory, processor, brand name and its colour which of this information is data. The messages and comments you make on social media platforms are also data. Even the voice calls in the video call to make her also data. Does we are surrounded by data everywhere it can be in any form such as text, number, image, video, audio and so on.

Example 2 :- When you search something on Google information like your location your device type internet service provider internet speed the search query you type your past search history and what kind of results reflect on previously information on what search results other people with similar search queries clicked on and finally the content of the web pages whose lives are shown as results are examples of some of the data that Google used to determine and show you the relevant results.

Example 3 :- When you are on Netflix it uses the data about the movie you may have watched her browse previously such as their genres, star cast, duration, language and so on to decide which movies to recommend you next. That is why if you have watched Golmaal Returns earlier it is like to show Judwaa 2, happy new year as next movies you could watch. Also as you watch or browse through more movies it continuously keep storing and updating the data to and uses that for future recommendations.

Example 4 :- When you book tickets on IRCTC and suppose you find that the ticket is in waiting list then it also shows the probability of a ticket getting confirmed. What data do you think IRCTC used to predict this. Yes as you may have guessed some of the data it is likely to use are the trains, past booking and cancellation history. Its current booking status and how long you have been on the waitlist at the moment and so on. You would notice that it keeps updating it predictions as more people book or cancel the ticket.

Similarly banks also maintain data about their customers such as customer's age, his occupation, how much he/she earns, where does he live, weather he has his own house or lives in a rented property, how much loan has taken in the past and whether he repay it on time or not and so on to determine the customers credit worthiness.

So from all these examples we can see that data is at the heart of machine learning and it is the data which enables machine learning to do all the amazing things it does. The amount of data produced captured and analyzed as seen an exponential rise in the last few years resulting in Rapid advancement in the field of machine learning and its applications. The more data you have the more patterns and insights you can derive from it and use them for a variety of tasks and to answer the number of questions.

Today billions of searches are made on Google everyday and billions of people watch videos on YouTube or like or comment on something on social media such as Facebook. Every single page you visit on internet, every click and scroll You Make, every email you open your send, every word you speak to any voice assistant on your phone is data which is produced stored and analyzed.

This data generation and capturing is not limited to just the online world anymore. If you use a fitness band for health parameters such as heart beat, body temperature, pulse rate and data on your physical activity is being generated and recorded continuously.

Similarly when you travel from one place to another and if you use Google Maps a lot of data is generated for your journey such as the route you took, your average time, stops made in between and so on. All of this data is captured.

As you can imagine the share amount of data that is generated and capture today is mind-boggling. This will continue to grow even further as data from billions of devices such as this lightning system at your home, CCTVs, the fire alarm, the cars, the refrigerator and so on will start getting captured we are the Internet of things in the future.

No wonder data is called the new oil and wood increasingly seen you battles emerging world over for the ownership of data and deriving benefit from it. Given the size of the data it is impossible for a human being to analyze it manually and Discover patterns and insights from it. That is why machine learning is becoming so popular and also making more and more accurate prediction.

Machine Learning Model

Ok now we know a bit about what data is and how it is power in machine learning. But how does one discover patterns and insight from this data to accomplish intelligent tasks. Such as making predictions or recommendations that we have discussed previously. This is the role of machine learning models or algorithms.

The way a computer program takes certain inputs, processes it and gives an output. Similarly a machine learning model or ML model takes the data as an input, process is it and comes up with patterns and inside. Let's call it intelligence as an output.

Now you can think of ML model as a black box which can analyze lots of data and come up with answer for a question we are interested in. Let see how?

Let take an example of a bank approving loan for customer and the question we want to answer is whether the bank should approve the loan for this customer or not. The bank has all the information or let say data about the customers such as his age, address, occupation, income, his bank balance and his deposit and withdrawal pattern. But how would a ML model look at this information to decide whether this customers should be given the loan or not. For this the model need to answer whether the customer will repay the loan on time or not. It is because if it doesn't repay the bank will loose the money it led to the customer.

So how will a machine learning model predict this? The machine learning model will analyze the data of all past customers to whom the bank lent money whether the repay the loan on time or not. And it will identify a pattern in this data as to who is likely to repay and who is not. Now of the customers information matches with the pattern of the customers have a good that is repay the loan on time his loan will be approved else the loan will not be approved.

So a machine learning model derives patterns and insights from past data and applies that intelligence on a new piece of data to make predictions for decisions about it. That in this article. In next article you will learn about different types of machine learning.

Attention readers! Don't stop learning now. Check out our articles to gain more knowledge.