It is my third time in San Francisco, and again I am highly impressed by the great architecture of the Golden Gate Bridge. This time I take a boat over to Sausalito at the other side of the bay. The museum there tells about the diverse history of the village and also, of course, the construction of the bridge. The statement and picture of Eddie Souza, a worker at the bridge, catches my attention.
“… That was the Depression and lots of unemployed men out of work were hanging on the fence of the bridge waiting for one of us get fired or hurt … “
What a tough job his must have been! All the workers who fought hard to do the work and not fall down from the bridge. They were the ones who built the bridge!
The stars in the spotlight
Later on I walk over the Golden Gate Bridge back to San Francisco. In the middle of the bridge I see the large memorial tablet that names all the relevant people who built the bridge. Do I see Eddie Souza's name at the tablet? No. None of the workers is mentioned here – only those who planned and directed the construction. I understand it would be hard to print the names of all workers who ever worked at the bridge here.
Credit to all who make the analytics life cycle run
However, those who have their names listed here could only shine because people like Eddie did their jobs. After awhile I think about my job and the many data scientists who often stand in the spotlight because they built a good, colourful and interactive model. So I wanted to say thank you to all my SAS colleagues and project members at customer sites who do an extremely important job so that we can build and present our models and that the analytics life cycle flows.
Machine learning models need data
We need data about our analysis objects to train the models. Database administrators store the data in their source operational systems. Those responsible for data integration access and transfer the data from different systems into data marts, data lakes that are accessible from analysis platforms. Data stewards and business experts make sure that the data quality is checked for criteria completeness, consistency, accuracy and timeliness. They profile the data from technical and business criteria to make sure that we can access it for analytics.
Artificial intelligence is enabled by computer software
So we need analytics software accessible on our computers. System administrators install the software and maintain it by applying updates and maintenance releases. Legal experts make sure that we have the appropriate software licenses and usage agreements to be allowed to use the software. System administrators also make sure that we can access the software with our credentials and only have access to that data we are supposed to use.
Business experts are our sparring partners
Every data has its history and its stories to tell. We need to understand the background and origin of the data. Otherwise we might build incorrect or irrelevant machine learning models. Thus we depend on the input and feedback from business experts who understand the business background of the data and know the operational process that generates the data. Business experts are also the ones who help us to calibrate the models so that they can be used in the operational process to make business decisions.
IT puts the models into production
The story must not end here. Just building the model is in many cases not enough. You must put the model into production so that it generates predictions and forecasts and feeds business decisions. When we come up with a fancy model with a high lift, good explanation of customer behaviour or precise demand forecast, we know that the success of the model will only materialise if colleagues from the IT department help us to put our models into production and integrate the analytics models into the production system.
AX2018: See you at the Analytics Experience conference in Milan!
Next time, e.g., at the SAS AX2018 conference in Milan, when you see me presenting results from a machine learning model created with SAS® Viya®, don't forget: I had help from others to walk through the analytics life cycle and generate the results that I am going to present.
I look forward to meeting you there and discussing your experience and requirements of building machine learning models covering the entire analytics life cycle.