Learn how to master the art of feature engineering

Feature engineering is the process of extracting and selecting relevant, informative and distinguishing characteristics from your data, which can subsequently be used as input for a machine learning algorithm. As the quality of your features largely influences the quality of the results, feature engineering is one of the keys to success in successfully applying machine learning. While it is typically a creative and work-intensive process, understanding the methodology, tricks of the trade and common pitfalls can help you go a long way.

Standard data mining/machine learning algorithms are provided ready-to-use by many libraries, toolkits and platforms, and can thus be applied by virtually anyone. However, simply running your data through the algorithm you have selected does not ensure that you will get good results.

The data needs to be presented in the right way to the algorithm. More precisely, you need to extract the most relevant, informative and distinguishing information from your data, a process that is called feature engineering.

Feature engineering is not only one of the most important steps in the whole data science workflow, it is also the step that will require most of your time as it is inevitably a creative, artisanal and trial-and-error process. During this iterative process, you will experience that it is possible to extract a wide variety of features from your data. Consequently, you will need to validate which features work and which don't, refine them or define additional ones, validate them again, etc.

Even though feature engineering is a trial-and-error process, there is a methodology behind involving some standard approaches, guidelines, recommendations, tricks of the trade and pitfalls, that you should be aware of and that can greatly help you in finding the most relevant features.

Are you interested in getting more insights and in learning more about the creative process of characterising data in terms of features that can help you reaching better results when you are applying a data mining/machine learning algorithm? Then register for our next training session dedicated to feature engineering. In this session we will discuss in detail the methodology behind it, including various methods for feature construction, selection, normalization, etc.