Extracting time components (month, year, weekday,...) from timestamp data

Timestamp (date-time) data includes many features that cycle over time periods: Hourly, daily, weekly, or monthly. Keshif offers the easiest and fastest way to extract, visualize, and analyze these periodic time features with simple, clear and interactive charts.

To extract a time feature (such as month, day-of-week. day-of-month, or hour): 

  • click "derive" button, and 
  • choose the time feature you'd like to extract from this data. 

You can open derive menu from a timestamp chart within the dashboard, or within the attribute panel when building a dashboard.


These extracted (derived) charts and data features work just like any other feature in Keshif. You can highlight, filter, and compare trends across months, weekdays, hours and day of month. In the example above, it seems that events between 0:00 (midnight) and 5:00am are more common on Sunday and Saturday, which reveals an expected "weekend" effect on the way officers are dispatched! 

Month chart: Auto-ordered from January to December in a logical way. You can choose to sort the months by value in chart configuration options.

Day-of-week chart: Auto-ordered from Sunday through Monday to Saturday in a logical way. You can change this fixed sorting order or auto-order by frequency in your data.

Day-of-month chart: This is presented as a histogram. The values range from 1 and 31.

Hour chart: This is presented as a histogram. The values range from 0 :00 to 24 :00.

Special Features

  • Keshif auto-detects which time features would be supported. For example, if your input is in date only (and does not include time of day), you'll not be able to extract "hour" summary.
  • You can delete an extracted feature from the dataset attribute list by selecting "Derive" option, and "Remove Derivation".


  • We recommend that you keep a single, fully noted time column in your data, and remove columns that may denote month, day, time separately. This makes the data easier to manage, helps you avoid data errors and processing errors. Remember that simpler is better! And, storing the time as a single, well-defined moment of time is the best way forward for many datasets.