In this video we introduce the topic of data engineering. Understanding your data is so vitally important. It is the raw material you use to create results. The common phrase 'garbage in, garbage out' (excuse my American!) summarises the ability to win or ruin data science projects and entire products with good or poor data. We will discuss common pitfalls and introduce steps to overcome them.
This video shows an example of using segmentation on categorical data and we'll also find that we have just derived our first important machine learning algorithm: a decision tree.
In this video we discuss segmentation, a very simple way of generating a predictive classification model given some data. We'll see later how this is the basis for a fundamental but very powerful machine learning algorithm.
Now we have a firm understanding of how business problems map to solutions we need to learn the techniques to deliver the solutions. This section introduces the basic terminology and concepts used in data science.
In this video we will talk about the problems encountered in data science. We'll also discover how it fits into a process, which you can used as a plan. Finally, we'll look at the impacts of a Data Science project which will help you avoid any common pitfalls.
This section introduces Data Science. It explains what it is and why we need it. We discuss some of the reasons for doing Data Science and provides famous examples from around the world.
How can you be sure that your models continue to work in the future? One simple method of testing robustness is to add some noise to the test data. From the response we can infer whether the model is stable or unstable. See this Python Notebook to find out more.
To be able to create models, you need to know how to compare them. In this python notebook you will learn how to compare models quantitatively. I.e. create summary statistics for your model's performance.
You should always try and visualise your data science. Especially when measuring model performance because this allows you to easily compare and spot errors with models. This Python Notebook shows you visualise model performance.
Often data is produced by a process that has some natural hierarchy. If you have a clustering problem where this is true, hierarchical clustering works really well. Find out more in this Python Notebook.
Case studies and industry analysis from our team. No hype, roughly monthly.