Embedded Files
Even after collecting and cleaning the data you may not have the data as required to meet your business case or answer the questions that have been posed. In this case, you will need to extract the correct features of the data, through Feature Transformation or Feature Engineering. Feature transformation is common in data-driven projects.
Feature transformation often takes the form of providing summaries of the data, normally averages for a given temporal or spatial range. For example, data may be collected every few seconds but the project requires hourly average to understand when productivity is highest and lowest. Another common reason to carry out feature transformation is to reduce the dimensions of the data and make analysis viable for a given resource. Dimension reduction will normally find functions of the data to use in the modelling process rather than the original data streams. A further reason to do feature transformation is to move between the time and spectral domains again reducing the dimensions of the data significantly.
Data Summaries
Data Summaries
Dimension reduction
Dimension reduction
Spectral analysis
Spectral analysis
Using the workflow - reporting on the features
Using the workflow - reporting on the features
Feature Reports
Feature Reports
Discovering, or engineering, informative features is a critical element of the DCM process, the nuances of which can be abstract to understand for an outside observer. Use this reporting template to record the key decisions and characteristics of the features generated throughout your project.
Page updated
Report abuse