Common terminologies used in Machine Learning and Artificial Intelligence
In this article, we’ll introduce you to various common terminologies used in the machine learning and artificial intelligence industry. Without any further delay let’s begin!
Note: If you are more interested in learning concepts in an Audio-Visual format, We have this entire article explained in the video below. If not, you may continue reading.
In order to explain these terms, I’m going to use a very simple chart-
On the horizontal axis. I represent “the value-added to an organization” while on the vertical axis, I am representing “the complexity of doing this practice”. Let’s dig in!
The first term you might come across is Reporting. This is at the lowest end of the Data Science spectrum. These are those kinds of reports which you can automate with minimum effort. These can include your daily, weekly, and monthly reports that use the same data sources and the same columns every time.
The next basic terms in this spectrum are MIS or Management Information Systems and Detective Analysis. MIS is used in organizations to track what is happening throughout the organization. Most of the time MIS answers the question “What has happened?” and the moment you start asking questions based on the data, you create a need to do a detective analysis. Detective analysis typically answers “Why has something happened?”
Here is an example of one such system for a bank sourcing business from several locations, as you can see here, a look at this report can tell you several things-
It can tell you which are the top cities or regions bringing in business for the bank. It can also tell you how much of this business is of bad quality. But why is this important? Well, this could be the business that the bank does not want to bring in. Right! It’s a very important metric. As I said before, it’s a very useful start for a business but by itself, the MIS cannot do a lot. This is where detective analysis proves to be very useful.
The next item you might have heard is Dashboard or Business Intelligence or BI. It is the utopia of the world. Imagine a world where every action about your business is reflected in front of screens of your business executives. That is business intelligence for you. Dashboards are used to answer “What is happening now in my business?”.
Here’s an example of a dashboard-
This is telling us about our user demographics. How many users visited the site? What time saw the most traffic? All of these questions answered in the near-realtime for the users of this dashboard. Useful, isn’t it?
The next item you might have commonly heard is called Predictive Modeling. What is predictive modeling?
Well, when you gather all the data to predict what is likely to happen at a granular level, this is called predictive modeling.
For example, for a bank, which customer is likely to default in the next 30 days is an answer which will use predictive modeling.
In order to do this, we will need to use past data about all the customers and look at the trends and reasons why customers default to predict the chances of failure of each customer we have today. Predictive analytics typically answers “What is likely to happen at a very granular level?”.
The next common term you might have heard is Forecasting.
Forecasting is a process of estimating the future based on past as well as present data.
It’s usually done at an aggregate level. For example, how many customers will fly on a particular flight, or how many calls can we expect in the next hour?
These are forecasting problems. You need an answer at an overall level for planning for your business. And this is how the outcome of a forecast would look like typically-
You see what is the expected demand, the upper bound, and the lower bound in this image. But if you need a more granular read, for example, which customers are likely to travel with the next flight? and which are likely cancel?, you will need to perform predictive modeling. Similarly, if you want to predict which customers are likely to buy a product in the next month? that will be a problem for predictive modeling. No outcome of a predictive model would look like this at a customer level. You will predict a particular outcome and then take actions accordingly.
Finally, one of the most common buzzwords. What is machine learning?
The machine learning is a system of teaching machines to learn things and improve predictions or behavior, based on data on their own.
For example, creating an algorithm that continues to become better based on consumer behavior is an application of machine learning, a classic case of a recommendation system, as you can see in this image-
Let’s take a moment to understand the difference between machine learning and statistical modeling. Take a look at this image from McKinsey-
It plots two variables in a graphical format. The slope line that you see here or the nonlinear line is classic regression analysis, plotting that line to understand the difference between the two drivers, which is A and B, or the two variables, in this case, is a statistical modeling task.
But when we see in the contours generated by the machine learning algorithm, we witness the statistical modeling is in no way comparable to the problem at hand. The contours of ML seem to capture all patterns beyond any boundaries of linearity or even continuity of the boundaries. This is what machine learning can do for you. Here’s a light-hearted take on how a statistician framed a model result as compared to a machine learning professional. Take a moment to read through this-
Do you see the subtle difference between the two?
Deep learning, a subset of machine learning is a very popular field these days and neural networks are deep learning techniques used for building models. Just to note here that the word deep comes from how many layers a neural network has or what is the depth of the neural network. This graph illustrates how well deep learning algorithms have proven to be-
The more data you feed the model, the better the performance becomes. Old algorithms plateau after a certain point, regardless of how much data you input, whereas deep learning models consistently improve. A few examples of deep learning applications are speech recognition, image recognition, natural language processing among a whole host of others.
And the final word in a spectrum is Artificial Intelligence. The applications of AI include chatbots and robots. There are a plethora of other applications in use today which fall into the bracket.
This article covered some common terminologies used in machine learning and artificial intelligence. I hope you understood the concepts explained in this article well.
If you are looking to kick start your Data Science Journey and want every topic under one roof, your search stops here. Check out Analytics Vidhya’s Certified AI & ML BlackBelt Plus Program
If you have any questions, let me know in the comments section!