From the course: Introduction to Artificial Intelligence

Common algorithms

From the course: Introduction to Artificial Intelligence

Common algorithms

“

- Machine learning is one of the most popular areas in artificial intelligence. That's partly because of the explosion of data but it's also because of huge advances in machine learning algorithms. Machine learning on its own is just a set of techniques. It's a way to build systems that learn through data. They move beyond the early AI systems that needed to be explicitly programmed. But there isn't one big machine learning program like Microsoft Office. Instead, you have many different machine learning algorithms. Most of these algorithms are borrowed from statistics. The key thing to remember is that each of these algorithms are like a chef's kitchen tool. Spoons are used for stirring and knives are used for cutting, but sometimes, chefs can find new creative uses for that tool. Some chefs use the side of their knife to crush a garlic clove or use a spoon to twirl spaghetti. I once worked for a company that was using credit card data to try and come up with customer promotions. They started out using supervised machine learning to classify the customers into two different groups. Remember, this is called binary classification. The first group of customers used promotions and the second group never used them. Then they used machine learning algorithms to train the system on this binary classification. Once they classified their customers into these two groups, they used unsupervised machine learning. They wanted to see if they could learn something about their customers who used promotions so they let the machine create clusters of this specific customer. Remember that unsupervised machine learning lets the system create its own clusters based on the patterns that it sees in the data. It's the same way that my son found clusters in unknown candy. What they found was that within this group who used promotions, there was an even smaller group of customers who always used promotions. They called them the promotions super users. Since this company was paid based on the success of its promotions, this was a great group to know. They found that this group used promotions for products, services, and restaurants. Those customers just love saving money, so the company tweaked the algorithm to offer more promotions to these promotions super users. That small change helped boost their overall success rate. Now, this organization used both supervised and unsupervised machine learning, so you'll have some algorithms that work best with supervised classifying and other algorithms that work best with unsupervised clustering. That's why it's important to see some of the machine learning algorithms that are available. Each one has its own strengths and weaknesses. Some take up more processing power while others are lighter and less accurate. Each algorithm is primarily used for either supervised or unsupervised learning, but there are algorithms that you can use for both. Like any good chef, the true creativity is not the tools, the ingredients. Your data science team might want to mix and match these algorithms to gain the greatest insights from your data. These machine learning algorithms are available in most of the machine learning software toolkits so it's here where your data science team can use its skill and creativity.

Contents