3 Replies Latest reply on Apr 8, 2019 3:00 AM by Jonathan Lee

    Outlier filter

    Jonathan Lee

      Hi,

       

       

      Great to be here.

      Currently analyzing 3 years membership data for a retail company, a few Millions rows

       

      Outlier is great, as we spotted new things.

      But now I am running market basket analysis,

      need to filter out many type of outlier for a more accurate recommendation for general customer

      (those who buy too much, those who visit too frequent etc)

       

      I could only think of running the flow in prep, then use tableau to manually circle and exclude the outlier,

      or as other post stated create calculated field to filter.

       

      It would be great if just a button we could exclude outlier in prep.

       

       

      Thanks

      Jonathan

       

      Prep Version: 2019.1.3

        • 1. Re: Outlier filter
          Ken Flerlage

          You'll need to put some definitions around what is an outlier and what is not. If you did that, you could easily create a calculated field that acts as a filter.

          • 2. Re: Outlier filter
            pedropablo.rodriguez

            Hi Jonatha,

             

            I recommend you creste with calculations different conditions with IF, in order to create new categories and determinate how tableau desktop use these categories as filters.

             

            could you please upload a set of data to help you better?

             

            Regards

            • 3. Re: Outlier filter
              Jonathan Lee

              hi,

               

              it's a suggestion in prep, just like how prep could exclude null value.

              as i am analyzing membership data:

               

              around 5% of member buy with out of normal high amount

              another 5% of member buy with out of normal high frequency

               

              I could exclude them in tableau desktop, but it would be great if there are a button to exclude them,

              easier than creating a box plot to locate below Q1 and over Q3 data

               

              thanks