Once you've identified the problem you need to solve and how AI can solve it, you can begin working with the data you'll use in the machine learning process. You'll need to prepare your data in various ways before you can build effective models from it.

By the time you have identified the business problems you need to solve and have collected, prepared, and analyzed the dataset, you will have a pretty good idea of how you might implement an appropriate model. At this point, you can start setting up and training a model, experimenting with it, and tuning it to produce the type of outcome you're looking for, at a performance level that meets your requirements.

Now that you've gone through the general machine learning workflow, you can start to apply that workflow to different machine learning algorithms. One of the most fundamental algorithms for predicting data is linear regression. You actually created a linear regression model as part of the earlier workflow discussion, but now you'll dive deeper into how this algorithm works and how it can solve your complex business problems.

There are many other algorithms that can perform regression analysis beyond just linear regression. Forecasting is one such alternative that excels at making numeric predictions from data that follows time-based patterns—everything from the weather to stock prices, and much more.

You've built regression models that can tackle numeric data, but that is just one common application of machine learning. Another major application is the act of classifying data. A data example is something, and simultaneously is not something else. In this lesson, you'll use common techniques for building classification machine learning models.

You've built models to tackle regression problems, forecasting problems, and classification problems. One of the other major machine learning tasks that you might want to engage in is clustering, a form of unsupervised learning. In this lesson, you'll see how a machine learning model can help you identify useful patterns even when the data you have to work with isn't labeled.

You've built supervised machine learning models from fundamental algorithms like linear regression, logistic regression, and k-nearest neighbor. These algorithms can get you pretty far in many scenarios, but they are not the only algorithms that can meet your needs. In this lesson, you'll build machine learning models from decision trees and random forests—two alternative approaches to solving regression and classification problems.

Another alternative approach to regression and classification comes in the form of support-vector machines (SVMs). In this lesson, you'll build SVMs that can do a good job of handling outliers and tackling high-dimensional data in an efficient manner.

All of the algorithms discussed thus far fall under the general umbrella of machine learning. While they are powerful and complex in their own right, the algorithms that make up the subdomain of deep learning—called artificial neural networks (ANNs)—are even more so. In this lesson, you'll build ANNs that can tackle the same basic types of tasks (regression, classification, etc.), while being better suited to solving more complicated and data-rich problems. These include problems in the domains of computer vision and natural language processing (NLP).

You've developed many different machine learning models throughout the course, but you've worked on them in isolation. In the real world, you'll need to incorporate your models into a wider context so that they can actually provide value to your business. So, in this lesson, you'll learn how to take the models you've created (and will create in the future) and put them into operation. That way, they'll truly be able to start solving your business problems.

Software development is an ongoing process, and machine learning is no different. Once you push your system to production, you cannot simply walk away and expect it to work flawlessly. Instead, you must put effort into maintaining that production environment, whether it's a simple deployment server or a more advanced data pipeline.

You'll wrap things up and then validate what you've learned in this course by taking an assessment.