Can machines take over humans? Answer is NO! Can machines take over daunting and repeatable mundane jobs from humans? Answer is Big Yes! And it is all possible because of Machine learning.
So, what is Machine Learning?
Machine Learning is a field of study under Artificial intelligence, which enables the machines to make decisions based on data models with minimal programming and manual interventions. So, what we do is something called “Training” the machines using different algorithms, data and automating them.
Before getting into why machine learning, let’s talk about numbers.
According to Indeed, The Best Job in 2019 is Machine learning engineer with an annual salary of about $146085 per year, with a growth rate of 344%. In India, the average salary of a machine learning engineer per annum starts at 8 LPA INR and goes up to 15 LPA INR.
So why machine learning?
Simple answer is to utilize the full potential of increase in data. We generate so much data every second that they can be used to create insights and draw patterns. That is what machine learning does! Machine learning uses the data and creates insights that are useful for the end user, with little manual intervention. Since it is data-driven, the insights gathered, and decisions made using Machine learning is pretty accurate.
Image Credits : https://lorilewismedia.com/
Knowingly or unknowingly we are using machine learning in our day to day life! Some of the examples are:
Google Maps:
Google maps have been the savior to many drivers for years now! The peace of mind you get, when you see your route to office is green is something everyone has experienced. Do Google maps know that there is a roadblock in this crossing, traffic jam at the toll or you need to take a diversion to reach your destination soon? Well it’s machine learning. Google maps have the data of the route, along with information from people using the route – such as average speed, location etc., trains its prediction model and gives us the optimal route and information.
Product recommendations:
Have you ever noticed this thing? You search for a yellow umbrella in amazon, same pops up in your Facebook feed, YouTube ads, in game ads while you play on your mobile and you visit any website, somewhere around the corner, you see the ways to buy yellow umbrellas!
Also, you get related items based on what you searched for. Say for example a holder for that umbrella. This is based on information from different users who bought the yellow umbrella that bought the holder as well. Google tracks its users’ search history and that data is used for prediction. May seem annoying, but you get best recommendations and user experience because of this feature! All thanks to Machine learning algorithms.
Netflix / Prime and other OTT platform’s movie recommendations:
“You may also like….”, this feature is what helps us find hidden gems of movies. Yes! That recommendation is possible with Machine learning! Not only suggestions are given based on what others watch, who watched the same movie/show as you, these recommendation engines actually analyses your data consumption patterns. See three horror movies in a row, and you will be watching more of it!
Social Media:
By now, you know the drill on how recommendations work. So, friend suggestions are based on the patterns just like movie and product recommendations. But there is one more outstanding feature in social media where machine learning algorithms are used. The Auto-Tagging feature. Based on the image recognition algorithms of machine learning, Facebook actually knows how you look and suggests that you are in that picture for easy tagging! How easy is that? Especially when you have a lot of friends, tagging them gets better & easier.
Personal Assistants:
Be it Alexa or Google Assistant or Siri, they work using various machine learning applications. They work based on a combination of speech recognition, language processing and speech to text conversions. Your personal assistant has various data loaded to it and trained already. So, when a query is made it recalls and gives output based on the prediction it makes. Life is surely simple with an assistant to remind us of our schedule & all mobile devices ensure that everyone gets the luxury of having an assistant. Utilizing it to maximum potential is in the user hands.
Based on the examples above, you can see a machine learning has the following process:
- Defining Objective
This is the very first step in any machine learning process. Define your end goal. For what this machine learning program is for. From above examples – to suggest a movie or product, to predict travel time or destination eta, tagging you in a picture etc.
- Data Collection
This is the foundation step. For future prediction, we need lots of historic data. Different varieties and types of data can be helpful in getting better predictions. Data collection is also known as scrapping of the data.
- Data Preparation
The data collected can be of different form, may contain inconsistencies and redundancies. In the Data preparation stage, these inconsistencies are removed. This is very important as duplicates and inconsistencies may result in wrong prediction.
- Data Analysis/ Exploration
Once the data is prepared, then an exploratory data analysis is made. In the data exploration stage, patterns and trends are obtained from prepared data. The data is also grouped based on their characteristics.
- Building A Machine Learning Model:
A machine learning algorithm is a set of rules or techniques to draw conclusions from data using a mathematical approach. Algorithm is the logic of the machine learning process. With an algorithm suitable for the particular objective, a model i.e. a representation of the objective is made.
- Training the model:
The cleaned data from step 4 is actually divided into two parts. One is input data or training data and other is reference data or testing data. The training data is given as input to the model and various results are observed.
- Testing the model:
Now with the help of test data/reference data that we have, we will test the model and fine tune its efficiency and accuracy. This will improve the overall performance of the model.
- Predictions:
The trained and tested model will be used for making predictions. The end result will be the objective that we defined in the beginning of the machine learning process.
Based on the data, the Machine learning process is divided into three types. Supervised learning, unsupervised learning and reinforcement learning. More about that and various machine learning algorithms will be presented in another blog.
Meanwhile, if you want to deep-dive and learn machine learning, new batches are starting soon! Our classes are instructor-led, and project based. Contact us to enroll in our Machine learning course.