DataRobot is amazing. It is gonna kick out many data scientists out of job, lol ! Great illustration Prof. Following your videos since the FinTech course.
I see the advantage and the time saving capabilities of this, however this is quite dangerous on the wrong hands. I foresee a lot of people just importing everything and trying to make a prediction. I would be quite curious of how the features were selected, were there any engineered features, how does it handle encoding, and how does it prevent overfitting. Plus I do like to keep my job and I do not want any software to substitute me 😉
Agreed. The modeler already knew which features to select and everything got used in every model even though attempting to reduce the dimensionality would surely have resulted in a better model. Also no strategy to balance the dataset. It was also a very clean dataset not realistic to the real world. I noted there was some missing value imputation but no information as to the calculation method used or comparison of the results from different methodologies. Also no information about getting real-time production data to the model, perhaps from multiple different platforms. Is there any PCA feature selection? Missing value imputation was referred to as feature engineering but no manufactured or concatenated features? API deployment?
Hi Rayan, Thanks a lot for your great session, just want to ask about the complete course on Datarobot, if you can tell me the name of the course on Udemy
DataRobot is amazing. It is gonna kick out many data scientists out of job, lol ! Great illustration Prof. Following your videos since the FinTech course.
I see the advantage and the time saving capabilities of this, however this is quite dangerous on the wrong hands. I foresee a lot of people just importing everything and trying to make a prediction. I would be quite curious of how the features were selected, were there any engineered features, how does it handle encoding, and how does it prevent overfitting. Plus I do like to keep my job and I do not want any software to substitute me 😉
Agreed. The modeler already knew which features to select and everything got used in every model even though attempting to reduce the dimensionality would surely have resulted in a better model. Also no strategy to balance the dataset. It was also a very clean dataset not realistic to the real world. I noted there was some missing value imputation but no information as to the calculation method used or comparison of the results from different methodologies. Also no information about getting real-time production data to the model, perhaps from multiple different platforms. Is there any PCA feature selection? Missing value imputation was referred to as feature engineering but no manufactured or concatenated features? API deployment?
Hi Rayan, Thanks a lot for your great session, just want to ask about the complete course on Datarobot, if you can tell me the name of the course on Udemy
Thank you so much.
Thanks for the illustrations
Excellent Work Sir
Awesome stuff :)
what to do if we have target leakage ?
Sir, what's the basic eligibility to become an ai researcher
understand me
I Want Data from Casinos and Crypto Markets so i can train an Ai and hit the jackpot lol
cool