Machine learning (ML) HADOOP system, based on PYTHON

Providing licensed software, primary code (C#, Python), and configuration files, serving as a basis for the future development of Machine Learning (ML) HADOOP.
Hadoop Data Trigger procedure:  The procedure is executed over a certain period of time (OPT), for example, initially 1 second. The OPT can be changed periodically by the ML system, which corresponds to the nature of the data( e.g. mean value to date).
1) Create an external table of Hadoop data from which the triggers are made. An outer table contains 1 to n triggers, representing a set of triggers. It is acceptable to have groups of such external tables that can be created by different sets of data. Each set consists of data in Hadoop with a single “date-timestamp” parameter, serving as Primary Key on the outer table.
2) Add a new row t), but in-memory (creates a multi-row table with the Primary Key “date-timestamps” parameter).
 3) Compare the last row added with the previous one. Creating a flag result from the comparison:
4) With “different” on a trigger, it activates the Stored Procedure (when a trigger definition is set, a Stored Procedure name is set to activate when the trigger content is changed). The difference between the two “date-timestamp” of the two rows (PM2P) is ouput to the ML system. The ML system consists of an interface with the Python Triggering Procedure (described above) which, through Python Managing Code, makes the connection to a Python library, for example: Scikit-learn, TensonFlow:
1)  New PM2P (obtained from the Procedure)
 2) Trained PM2P (obtained from the Python library)
 3) OPT={Average | (Trained PM2P / 4 | …}
4) The OPT is transmitted to the Procedure
Building the configuration for “OLTP in memory”.
Configuration steps:
1) Identify the “in-memory” table
2) Create “Columnstone Indexes” for each trigger
3) Connect the SQL Server 2017 Services to the “in-memory” table

The price of an annual licence is 2000EURO without VAT.