[DOCS] Added Data Frames subsection to ML section.

docs/en/stack/ml/dataframes.asciidoc

@@ -0,0 +1,57 @@
+[[ml-dataframes]]
+=== {dataframes-cap}
+
+{dataframes-cap} feature is available in 7.2 and later.
+
+A _{dataframe}_ is a transformation of a dataset by certain rules defined during
+the creation of the {dataframe}. You can think of it like a spreadsheet or a 
+data table that makes your data ready to be analyzed and organized.
+
+{es} datasets consist of individual documents that have fields and
+values in each field. This architecture makes search easy but on the other hand, 
+makes it hard to run analyses that require reorganized or summarized fields of 
+the dataset. {ml-cap} analyses need clean and transformed data and that is the 
+point where {dataframes} come into play.
+
+To transform the data into a {dataframe}, you need to define a _pivot_. During
+pivoting, you create a set of features that transform the dataset into a
+different, more digestible format to make calculations on your data. Pivoting
+results in a summary of your dataset (which is the {dataframe} itself).
+
+Defining a pivot consist of two main parts. First, you select one or more fields 
+that your dataset will be grouped by. Principally you can select categorical 
+fields (terms) for grouping. You can also select numerical fields, in this case, 
+the field values will be bucketed using an interval you specify. The calculation
+will run against every bucket that was created this way.
+
+The second step is selecting one or more aggregations to perform calculation over
+the dataset. When using aggregations, you practically ask questions about the 
+dataset. There are different types of aggregations, each with its own purpose and 
+output. You can learn more about the supported aggregations and group-by fields 
+here (!add a link!).
+
+As an optional step, it's also possible to add a query to further limit the 
+scope of the aggregation.
+
+IMPORTANT: In 7.2, you can build {dataframes} on the top of a static dataset. 
+When new data comes into the index, you have to perform the transformation again 
+on the altered data. Using {dataframes} does not require {dfeeds}. 
+{con-dataframes-cap} will be introduced in a later version.
+
+.Example
+
+Put the case that you run a webshop that sells clothes. Every order creates a 
+document that contains a unique order ID, the name and the category of the 
+ordered product, its price, the ordered quantity, the exact date of the order, 
+and some customer information (name, gender, location, etc). Your dataset 
+contains the documents of all the transactions from last year.
+
+If you want to check the sales in the different categories in your last fiscal year,
+define a {dataframe} that is grouped by the product categories (women's shoes, men's
+clothing, etc.) and the order date histogram with the interval of the last year, 
+then set a sum aggregation on the ordered quantity. The result is a {dataframe} 
+pivot that shows the number of sold items in every product category in the last 
+year.
+
+IMPORTANT: Creating a {dataframe} leaves your source index intact. A new index will 
+be created dedicated to the {dataframe}.

docs/en/stack/ml/overview.asciidoc

@@ -11,3 +11,4 @@ include::buckets.asciidoc[]
 include::calendars.asciidoc[]
 include::rules.asciidoc[]
 include::architecture.asciidoc[]
+include::dataframes.asciidoc[]