Machine Learning Model Migration

AREA:

MACHINE LEARNING, IT INFRASTRUCTURE, DEVELOPMENT SUPPORT

PROJECT LENGTH:

4 MONTHS

BRIEF

One of Hungary’s leading financial institutions decided to introduce new machine learning-based solutions at group level. The first implementation phase of which was to introduce the solutions at the bank located in Hungary.

The aim of the project was to create a machine learning model that makes customer segments easily definable and track their changes through a self-learning process.

The BCS Consulting team provided assistance in the following parts of the project:

Migration of the model developed in an AWS cloud environment to an on-premise environment (together with the Vienna headquarters)

Creating and operating an on-premise environment: the project had to take into account the banking operating environment and the guidelines of regulators

TECHNOLOGY:

R, RSTUDIO SERVER, GIT, BITBUCKET, ORACLE DB, LINUX, NEXUS, ZABBIX

PROCESS

1.

Assessing requirements: how much storage space, how much memory, what software and GUI are needed by developers?

2.

Assessing opportunities: What is the IT infrastructure and what entitlements are needed?

3.

Realization
i. Creating an IT architecture: Creating servers, installing programs
ii) Ensure access to RStudio Server development environment from computers used in development (firewall rules, etc.)
iii) Provide database connections to the model (firewall rules, connection strings, technical users, etc.)
iv) Provide version management with Git via Bitbucket repo
v) Test database connections and version management through RStudio Server
vi) Domestic migration of the Viennese model
vii) Conversion of the Vienna model to the domestic data scope (database connections, bringing input data into the correct form)
viii) Create a scheduler script
ix) Develop encryption for technical users and passwords
x) Log management design (log rotation)
xi) Monitoring development
xii) Release Structure and Release Process development
xiii) Creation of developmental and operational documentation

4.

Testing
i) Continuous testing of functions during development on the developer server (DEV)
ii) First Release testing in the Integration (INT) environment

5.

Deploy to production server

RESULT

A machine learning model and environment created that makes it effortless to define customer segments and track changes in them using a self-learning process.