The Machine Learning Magic! Let's predict things!
In Part I, we have discussed different maintenance strategies, what is predictive maintenance and intoduced Azure ML.
Part II covered the steps involved in getting Azure ML up running:
In part III of this blog post series we establed an Azure ML model for predicting the type of vehicle breakdowns (electrical, mechanical or no break-down). In this post we will finally make use of all our preparatory work!
So let's assume we have some Azure ML competencies who has done all the setup in Azure ML, uploaded the known dataset, selected the most appropriate algorithm, trained it and scored it.
And let's assume we have some maintenance managers who would like to make use of this model.
Now, one approach is of course just to give the maintenance managers access to the Azure ML Studio and allow them to start to upload data and expose that data to our model. This is probable way above what we can realistically expect a maintenance manager to do!
Instead we can simply expose our model as a Web Service which can then be comsumed from "anything"! An obvious example could be from their favorite Enterprise Asset Management system where they probably already have some dashboards where our Web Service could be easily intergated.
When exposing a model as a service, Azure ML takes care of everything, including:
- Documentation describing REST call, input and output formatting
- Sample client code in C#, R and Python
- Deployment to Azure cloud
- Web service staging
So with all this we achieved a service where a maintenance manager could input a query like:
Should we take predictive action before to avoid a mechanical or a electrical breakdown?
Pretty cool I would say! Feel convinced Azure ML might posses some power that could provide value to your business?
PS: Credit to Tomas Grubliauskas for providing all the hardcore background material for these posts!