Incorrect predictions by classification

Use this course to bring your thinking about prediction errors to a practical level. With prediction by classification different types of prediction errors have different practical implications. Understand these differences to better build and assess the relevance of your predictions.

Understand the Trade-Offs between Prediction Errors

• Learn what is classification and the reasons for using it.

• Recognize the use of classification for prediction in technologies around you.

• Understand classification to predict and why prediction errors appear.

• Learn to think about the practical implications of prediction errors.

• Learn when aiming for no prediction errors is not a good idea.

See the Practical Relevance of Predictions

Technologies that use classification for prediction include E-mail spam filters, systems to detect suspicious purchases with your bank card and self-driving cars. Data scientists and their colleagues work with large amounts of data to make these technologies successful. This course aims at developing your intuition on the practical relevance of predictions. The absence of programming examples and mathematical detail is a key feature of this course.

The practical implications of classification prediction errors tend to receive no or little attention in courses or textbooks discussing machine learning classification methods. This course focusses on those implications and provides guidance on what to do when knowing them.

The course starts with defining classification and provides simple examples. It explains statistical classification and that uncertainty adds complexity to classification problems. It provides reasons for why people use classification methods and prediction is the reason that is central to this course. You will learn the main ideas about predictions and methods to make them.

Six different real-life applications of prediction by classification are around us or will be around us in the not-so-distant-future. You will understand the characteristics they share with each other and two of the applications will be discussed in detail throughout the course.

You will learn also why the presence of prediction errors is to an extent inevitable and how to use knowledge about their practical implications to your advantage. In addition, you will learn about some practical aspects in prediction projects related to prediction errors. One of them is the issue of setting almost zero prediction errors as your goal.