Data analytics jobs are on the boom while, at the same time, graduates of data science grow very quickly through traditional and online courses.
The data analytics tasks summarized as shown in CRISP-DM (see picture) that require technical expertise are partly being automated by tools such as WEKA and TPOT.
Will data analytics jobs be fully automated? Which data analytics tasks are difficult (or impossible) to automate?
An interesting question.
I think the most difficult parts would be at the beginning and the end of the process:
- To understand the context, i.e., the business understanding phase in CRISP-DM;
- To interpret the outcomes from the modelling phase, given a certain context of the business under consideration, for example, how to explain the features of each group from the k-means clustering.
The modelling part can be automated to a certain extent, like using a library, where algorithms are treated as various black boxes - data-in and data-out.
In brief, anything that can be treated purely as numbers - no matter what the numbers indicate - could be automated or programmed - again, thinking about what a library does.
Hope this makes sense. Thanks.
There is a quote "If I had an hour to solve a problem I’d spend 55 minutes thinking about the problem and 5 minutes thinking about solutions." It is often attributed to Albert Einstein but there is no evidence that he ever said that.
Regardless, the biggest challenge is Business understanding or Defining the problem, so it could have a practically feasible solution. This task cannot be automated. The next step is deciding what algorithm or method is the most efficient to solve the problem followed by the data that are required to feed the algorithm and then test it.
The bottom line: in data analytics one should never start with data just because some data are available. One should start with the specific statement of the business problem.
- Badges
- Science topic