^{Daniel Grecoe - A Microsoft Employee}

The code in this repository sets up different versions of an Azure Machine Learning Service.

For detailed information on the Azure Machine Learning Real Time Scoring Service deployed in this example read this document.

A real time scoring service is a REST endpoint in which consumers of the endpoint send in individual records to be scored synchronously. Applications for such a service can be found in predicitive maintenance or other applications where some immediate action should occur.

For example, a model is deployed for in a factory setting. At a given cadence the manufacturing equipement transmits it's current status and potentially the status over the last time period.

The real time scoring service would then predict if the machine is about to experience some sort of issue. Business logic built with this endpoint could then shut down the machine or schedule a maintenance ticket.

For detailed information on the Azure Machine Learning Batch Scoring Service deployed in this example read this document.

A batch scoring service is an asynchronous service that can be run either on a schedule or when triggered. Applications for such a service can be found in financial services or other applications where periodic scoring of large numbers of records are required and immediate action is not neccesary.

For example, certain fraudulent activity occurs over a period of time but is not detectable by individual transactions. Using a batch service, transaction historry would be collected by a back end system and placed in a storage area (many storage options are available for both input and outputs).

When the input data is collected, the batch service would then be triggered to predict fraud over some period of records and output the results into another storage location for further business processing.

Either path you choose you need to set up a conda environment to execute the scripts in as required Azure and Azure Machine Learning modules need to be present.

1. Clone this repo to your machine
2. Create the conda environment
   • conda env create -f environment.yml
   • conda activate SimpleModel
3. Choose the path you want to follow:
   • Real Time Scoring - Navigate to rtsreadme.md
   • Batch Scoring - Navigate to batchreadme.md

Regardless of path you execute, upon completion of either the rtscreate.py or batchcreate.py a log file directory is created with the following: