Quiver | Calculating KPIs for Time Series Data in Palantir Foundry

[Music] thank you [Music] today we'll be covering quiver's time series functionality and its ability to calculate kpis Based on data we'll use publicly available climate data from across the United States and try to determine which region of the country has had the most temperate climate over the last 20 years to do this we'll calculate the standard deviation of temperature over time in different regions using quiver's transformation tools let's get started in our current setup we have a Time series data set representing climate data from each of the climate divisions defined by the National Oceanic and Atmospheric Administration this includes temperature precipitation and drought risks data spanning more than 100 years to start let's add climate time series objects to a new quiver analysis we built these objects beforehand using our climate data set you can see that this object type has an icon next to it indicating that it contains time dependent properties we'll add it to quiver here we can see all the time series we have available they are broken down by state division within a state and the measurement they represent for our purposes today we're only concerned with average temperature so let's filter out objects with other measurements to do that we'll click objects then filter object set and add a filter where the element name is average temperature this shows that we now have a few hundred time series remaining let's get a sense for what these look like to view a subset of them click Time series then group time series plot in the right configuration panel choose the correct time series column we'll also change the number of series to show up to 10 plot them in different colors and show the division name for each we're now seeing a sample of the temperature data sets we can see that they are cyclic rising and falling each year which makes sense for temperature data this is showing the full data set which spans more than 100 years however we're only interested in the last 20 years so let's crop our view foreign shared time access which defines the time range that we see in the graph if we add it to our canvas we can easily adjust the date range we now see a more detailed view of the sample 10 times series shown here now let's calculate how temperate each division is we'll do this by calculating the standard deviation of mean temperature values for each division over the last 20 years to start let's create a transform table for the temperature time series objects we'll now see a list of all relevant objects to which we can apply transformations first we'll click the data to only look at the last 20 years click add transformation then under time series find the time slice transformation option this will keep only the data found within a certain time interval give the output column a name select the original time series data and choose the interval we defined in the graph above we can also Define a different time range by creating a new parameter we can now see a new column in the table showing previews of the clipped data next we'll add the standard deviation transformation this is found as a Time series numeric aggregation give the new output column a name then select the new time series column we just created and last choose standard deviation as the aggregation type this calculates the standard deviation of mean temperature for each division over the last 20 years let's clean this up a bit since there is a lot of information that we don't need we'll hide and delete unnecessary columns lastly we can sort the Visions by standard deviation in ascending order to find those with the smallest temperature swings throughout the year foreign are the most temperate in our data set while California and the Pacific Northwest also have moderate temperature swings but what if we only wanted to look at certain States maybe just in California all we'd have to do is add an additional filter at the start of our analysis return to the object filter and add an additional criteria for California the analysis is now automatically updated we can see new graphs of the time series data and our Transformations only include California divisions there's much more we could do using quiver so I encourage you to explore and find out how it can help you learn from your time series data at scale [Music] [Music] foreign

Channel: Palantir Developers

Views: 2,563

Keywords: palantir, PLTR, foundry, Quiver, FoundryQuiver

Id: x5m6DjijAPI

Length: 6min 1sec (361 seconds)

Published: Tue Mar 21 2023