Our software has many options available to create new features from the original features in your dataset. These options can be accessed by hovering over a feature either in Spreadsheet View or in the Feature List, and then right-clicking (desktop) or pressing B or Y (VR).
For all feature types, you will have the following options:
- Rename – Choose a new name for your feature (Note: When working with network data, this option will not appear for the Node ID column)
- Histogram – Display a 1D Histogram to see the distribution of the feature, similar to the one in Spreadsheet View
- Copy Name – Copies the name of the selected feature to the clipboard
- Copy All Names – Copies the names of all features to the clipboard (in the original order that they appear in the dataset) (only from the Features Panel)
- Copy Column – Copies all the row values of the selected column (only from Spreadsheet View)
Additional options are available depending on the type of the feature selected.
For Numerical features, the following features creation options are available:
- Convert to Categorical – Copies the selected feature and casts the new feature as Categorical
Normalize– Creates a new feature by applying the selected normalization to the selected feature:
- Normalize [0, 1] – Reorders all values in the original feature between 0 (for the lowest value) and 1 (for the highest)
- Log10 – Normalizes the original feature (see Normalization)
- IHST – Normalizes the original feature (see Normalization)
- Softmax –Normalizes the original feature (see Normalization)
Functions– Creates a new feature by applying the following functions to the selected feature:
- 1/x – New feature contains cells created by applying the function 1/x to the selected feature’s cells
- Binary Median – New feature contains cells with “below median” or “above median” depending on the values of the original feature for that row
- Above/Below Zero – New feature contains cells with “Above Zero” or “Below Zero” depending on the values of the original feature for that row (note that zeros will be included in the “Above Zero” category)
- Quartiles – Assigns quartiles to each row depending on the values of the original feature for that row
For Categorical features, you can create a new dataset by pivoting on the selected feature. Each row in this new dataset will be a unique value in the feature on which you choose to pivot. The columns will be calculated by applying the desired function (Min, Max, Mean, Median, Std, Sum, or All of the above) to each of the numerical features in the original dataset.
You also have the option to Convert to Numerical, which copies the selected feature and casts the new feature as Numerical if the original feature was Numerical Categorical, and results in a column of missing values if the original feature was String Categorical.
For Datetime features, the details shown above can be extracted from the feature. On Date features with no Time specified, extracting time will result in 12:00:00AM. On Time features with no Date specified, extracting Date will result in the current System Date. Please see the Accepted Date Formats page to check if your typical format is supported.