Introduction
Machine Learning (ML) is a subfield of Artificial Intelligence (AI) that provides systems the ability to automatically learn and improve from experience without being explicitly programmed. It focuses on the development of computer programs that can access data and use it to learn for themselves. The process of learning begins with observations of data such as examples, direct experience or instructions. It will then look for patterns and data and make better decisions in the future based on the examples that we provide. The primary aim is to allow the computers to learn automatically without human intervention or assistance, and adjust accordingly.
Classification
Source: https://www.youtube.com/watch?v=R_5BTa8PVdg
One of the main uses of ML is classification. As you can see from the video, there is a simple classification method for onions. In this case, the classification is made up of those two rods and essentially it is dividing the onions based upon their size. However, classification can become much more complex.
Source: https://www.youtube.com/watch?v=7o6iWffuuTQ
Classification is the process of predicting the class of given data points. Classes are sometimes called labels or categories. For example, in the coin video, the classes are 50, 20, 10, 5, 2 and 1. The task is to categorise our coins or sort them in those classes so we get the 50s together, the 20s together, etc. Another example might be spam detection, spam detection in the email service providers can be identified as a classification problem. This is a binary classification since there are only two classes and email can be spam or not spam. Classifier utilizes some training data to understand how given input variables relate to the class. In this case, spam emails and non-spam emails have to be used as a training data. When the classifier trains accurately, it can be used to detect an unseen email.
Classifiers
Classification is extremely useful and we can use it for various applications. For example, we have a set of pets and we can use their data features. So what are the elements which describe them? Maybe it could be their size or their color or the shape of their hair or even their tail, and then we can train an algorithm to recognise them.
Finally, the algorithm will learn how to place them into specific categories and then we can use them for different processing.
The AI system, which is used to classify this data as normally known as a classifier. So, in the pets example, a ‘1’ would mean that it has that feature and the ‘0’ would mean that it doesn't have that feature. Some of the features will, of course, fire or activate, and that means that that particular animal belongs to that class of animals. A high score, would means that it belongs to one class and the low score would mean that it belongs to another class. For example, if I have a class of small animals and they have pointed ears which are short, then most probably it belongs to the class of cats. And this process continues for each and every animal. There are different types of classifiers, such as logistic regression, support vector machines, naïve base and neural networks.
How does Classification Work?
At this stage, you might be wondering how classification actually works. For instance, if one wants to categorise cats and dogs, one would create an algorithm which manages to recognize whether an image represents a dog or a cat. The first thing we do is to extract the number of features, for example, size, colour, shape of ears, etc.
Second, we need to map them into a two-dimensional space. So, we divide that space based upon the features of those animals.
Third is to train the algorithm. In several cases, the algorithm manages to recognize all the cats and all the dogs. This is called overfitting. Overfitting occurs when the algorithm manages to recognise all the objects within our training set, but of course, only those objects. So, imagine we have a new picture of a cat, something which the algorithm never saw, and we would like to show it to the algorithm. In that case, the algorithm will most probably fail and it will fail because the algorithm learned how to recognise only those images in the training set. However, we want the algorithm to learn to recognise all existing cats and dogs, and not just those present in our training set. So, when choosing, a model, we need to choose something less accurate, but which manages to generalise better. By doing so, if we feed the algorithm an unseen image, it will probably recognise it and classify it in the correct group of animals.
Machines that Learn: The Amazon Case Study
AI, and in particular ML, are being increasingly used by businesses to enhance their operations and subsequently offer better services.
Source: https://www.youtube.com/watch?v=2DtyjC0UxTw
Amazon is using AI in practically everything! But did you ever wonder how your shopping experience on the Amazon website actually works? Which part of Amazon is actually learning how you are interacting with the website and giving you back information which is relevant to your search? Did you ever stop and wonder how does Amazon recommend that particular item which you are interested in? In fact, it is fair to say that Amazon excels in the way in which it uses AI.
So, how does Amazon help you locate the items you want to purchase? Sign in on https://www.amazon.com/ and then have a look at the different components which exist on the website. Try to locate those parts of the website which are actually offering recommendations.
The first thing you'll probably notice is the section which offers you personalised recommendations. This is a page full of products recommendations just for you. In fact, they have been customized based upon your past experiences. Amazon recommends a range of products from different categories which you have been browsing and places products that you are likely to click.
You'll also probably notice something called ‘Frequently bought together’. This recommendation has one main goal to increase average order value. So, it's not just about buying one product, but you might also want to buy an additional product which is related to that particular product. Recommendations aim to up-sell and crosssell products by providing suggestions based on the items in the shopping cart or based on products you’re currently looking at on the site.
Another relevant section is entitled ‘Customers who bought this item’. Over here you can see items which were bought together with the item which you actually bought, and Amazon is recommending that you also buy them. The company displays items that have been purchased together in the past with the goal of increasing average order value through up selling and cross-selling. These items are purchased together a little less often, then frequently bought together and is a way for Amazon to sell items that are not as popular to help retailers move their inventory.
The ‘Inspired by’ section is used to show items which are related and might be interesting. Amazon looks at products you've been browsing and recommends very similar products of different shapes, sizes and brands to help you find something very similar to a product you've already shown an interest in. It shows you different brands, colours, shapes and sizes with the hope that you will proceed to buy.
In the section entitled ‘A newer version’, Amazon shows you a new version of an item which you already own. If for instance, I look at an old e-book reader I bought from Amazon, there is a recommendation underneath the invoice letting me know that there is a newer version of the product that I can upgrade to. It's almost like a replenishment campaign, but for electronic devices.
Recommendations based upon previous purchases look at what you already bought and suggests alternative items that you might need, based upon that purchase. For instance, after purchasing an e-book reader, Amazon recommends a variety of different covers and cases for the exact product I had just purchased in an attempt to encourage a second purchase with a highly relevant cross-selling offer.
Best sellers form a specific product category which help people find popular products and buy from new categories that they may never have purchased from before, which opens up a whole range of up selling and cross-selling opportunities as well.
These are real life examples of how AI is being used in practice. In this case, Amazon is using it to recommend products. AI is basically crunching all the data about what you do and even what others do on the Amazon website, and presenting recommendations based upon the data.