Hello, I am a student of Data Science and this week I was offered my first job in the area. The job is in a consulting company to work with a partner that has bought the software DataRobot and they need a team to operate it.
I never heard about this software before. Did some research and apparently this software do the job of a data scientist. So I am afraid that I will be a monkey pressing buttons and not really work applying my data science skills. What do you guys think? Am I wrong?
I have been using Eureqa (academic license) to forecast stock market volatility. I was able to show that symbolic regression as implemented by Eureqa/DataRobot is able to outperform (provide a more accurate out-of-sample volatility forecast) the HAR model.
Is there a way of getting Eureqa to report the p-values for the models that it evolves? Is this functionality (p-values corresponding to models found by symbolic regressions) present in DataRobot's software?
Congrats on the job offer! I would echo what the previous posters have said and I'd definitely add that using DataRobot does not make you into a monkey button pusher! I promise. Having some fundamental knowledge in statistics/machine learning will certainly help you understand what DR is doing for you automatically and allow you to move faster.
You will still need to frame business problems correctly so that an algorithms can solve them.
You will still need to prep your data correctly.
You will still need knowledge around how to evaluate your best performing models.
And you will still need to know about refreshing your model given new dynamics in the system.
Keep studying data science and use DataRobot to accelerate your machine learning! 😄
Etc. Etc. Etc.
@Thodoris gave a great [quick] history lesson on what was once manual tasks
Hey there tecchUSA,
First of all, congratulations on your first job offer! May you have a wonderful career.
To answer your question, I think that we should look at the history of ML and put things into perspective:
Not long ago, libraries like SK-learn which provide you 1 line model training did not exist. This meant that people and researchers, would have to write their own models from scratch. Did the emergence of technologies (like SK-learn) make their jobs more boring? I would argue no. Once you debug your own random forest once, you do not want to do it again :).
AutoML is just the next step to making boring/repetitive tasks easier. This does not mean that Data Scientists become “monkeys” in any way. It means that Data Scientists have more time to define the problems correctly and work on what actually needs their attention.
Furthermore, you might find splitting your data into train/test or applying multiple models and preprocessing techniques interesting now, but you will lose interest in that sooner or later as it will get repetitive. Trust me, I have been there!
The world of Data Science is vast and modeling is just a small (but time-intensive) part of it. Do not be afraid of DataRobot. Embrace it and enjoy the comforts modern technology can provide you while focusing on the things that matter most!
I hope this answers your question tecchUSA, have a good day!
This is an extremely common question.
almost exactly the same asked on saturday https://community.datarobot.com/t5/research-center/is-data-robot-moving-a-data-scientists-cheese/m-p...
And last week https://community.datarobot.com/t5/automated-time-series/what-are-your-thoughts-on-datarobot-and-oth...
Check out the responses - basically the fact its in the process of being automated just indicates its a growing area. There is some need for data science understanding even when using automated tools - history in other related areas indicates that demand will only grow.
So actually, now is a great time to get involved - enjoy your studies, and your first job! And feel free to use the community for any more queries you have around the software or the field in general. Or even if you just want to share something interesting