Capabilities

Code Free and Production Native Data for Machine Learning & Analytics

Windocks code free feature data is a collection of database and table level transforms to source data from one or more databases, with data cleansing, normalization, rich statistics, and features.   The Windocks solution offers a new approach to feature data:  

  • Production native. The Windocks feature data pipeline is a repeatable, high performance solution that delivers feature data natively on the source data platform (Oracle, SQL Server, Snowflake, etc.).  Features are production ready, simplifying model retraining and A/B testing.

  • Code free feature data development. The Windocks web UI simplifies data sourcing and transforms, without SQL or Python coding. Windocks enables features to be developed by anyone, in hours, compared to what typically requires days of coding.

Anyone with analytical skills can now be productive producing robust feature data.  Windocks does not trade-off ease of use for dumbed down data, but goes beyond other solutions with advanced support for joining disparate data sources, enabling multiple and larger data sets, and other capabilities. 

Get a demo
Windocks Feature Data 1407x910

More on Windocks Feature Data 

Sourcing data from databases and warehouses

Windocks simplifies sourcing tables and merging columns from related tables from complex databases.   One option is to begin with subsetting the database or warehouse to reduce it's size, and focus the sampled data on the desired timeframe or by other attributes.  Windocks presents the collections of related tables available, and provides a simple method to join columns from related tables, to produce a data set needed. 

Cleansing, Normalization, and Statistics

Once a table is created that contains the needed source data, work can begin on cleansing and normalizing the data.   Again, this is a simple process, or dropping rows with null values, converting string data to numerical columns, and other steps.    At this stage aggregates are created, and rich statistical values can be computed (means, standard deviations, etc.).  

Features

The previous steps may have produced desired features, but in many cases the features will be produced by operating logic on the various statistics and row values.  Windocks includes simple to use formula logic to add features (similar to Excel logic). 

Run the transform and edit as needed

The above steps define the feature data transform, are saved and then "run."  The result is the delivery of the feature data table on the target database instance.  The data can also be delivered in the form of a flat file for further analysis.  Windocks transforms are easily edited, for updates and optimization.  Transforms are also naturally shared as well.  

Explore More Concepts

 

 

Let's chat!