How to set up a Machine Learning project in Azure

Posted by Daniel Bassett on Feb 4, 2020 2:45:27 PM

Machine Learning is a great tool for making predictions on company data. We have worked with companies and solved their problems including whether a customer or employee will leave their organisation, whether a subcontractor will go over budget, whether a retailer will vacate an asset, or determining financial records that are applicable for tax exemptions. Despite such obvious advantages often business managers don’t know where to start or how to implement a machine learning solution within their business. The Azure Machine Learning (Azure ML) service is a great solution for such people and offers a simplified way to start bringing machine learning into their organisation. It offers the ability to build machine learning pipelines within a graphical user interface that involves no coding and that connects easily with other Azure services such as Data Factory, to help you automate the entire process. Also, it offers advanced machine learning tools for existing data scientists and is highly scalable and able to adjust to increasing data loads. Thus, it can progress with your business as your machine learning capacity evolves. This makes Azure ML an ideal entry point for businesses wanting to implement machine learning, but who do not know how to get started or who are wanting an easily managed machine learning service. This is especially the case if you are already an Azure cloud-based customer. 

We often hear from businesses that they do not want a solution that is overly complicated or is a black box to the business. They generally want a solution that numerous employees are able to use rather than a select few. Azure ML goes a long way to offering this solution.  In the example below we built a machine learning pipeline with no code using the designer component of Azure ML. We are able to take in customer data, train a model that predicts whether a customer will churn (with > 80% accuracy), and then output that data into a file that is accessible to all within a business. Although this was all achieved with no code, if coding is needed to tackle a problem, you can add a script written in either Python or R within the pipeline. Thus, you only code when absolutely needed.  

Once we had the machine learning pipeline built, we then linked this to an Azure Data Factory. This is another service within Azure that helps us to automate the entire machine learning process. Once complete the whole machine learning project could run from beginning to end without us ever having to touch or think about it. The end result was that thousands of customers could be added to a folder, and seconds later the machine learning model would make predictions on every customer and then add this output (customer and their churn score) into a file or data store that can be accessed by anyone in the business. Obviously, this same scenario could be applied to any of the use cases we mentioned or any challenge that your business is wanting to solve.

On 13 February 2020 we will host a webinar giving a live demonstration of exactly how we achieved the above and how this can be applied to similar challenges within your business. If you are interested in using Azure ML services within your organisation, or know of anyone who might be interested, please join us for our webinar on the 13 February. 

Click here to register

Topics: Machine Learning