I'm a bit late to the party! 2 years since the videos were released and they're still proving to be massively helpful. I've enjoyed all the videos in the series so far. Most of all, the extra information you provide is priceless. Thanks a bunch. It's so useful to see you stepping through the process.
@raven53353 жыл бұрын
Your work is awesome.. thank you Abhishek
@ravulasandeep25123 жыл бұрын
Thanks for this initiative.Gaining a lot Abhishek.
@2mitable2 жыл бұрын
learn lot From you thanks for this playlist
@saravanakumarg1403 жыл бұрын
Great bro
@chienlingliu2 жыл бұрын
great video. What does the step_1.b.check() thing do ? is it a function you defined ?
@nisharnbhana4513 жыл бұрын
Hello, I am curious would you be able to explain how you record your videos (Screen recording &camera)? It seems that you can switch between computer screen and iPad and am wondering what software you use
@grow-with-abi3 жыл бұрын
Sir, for a classifier model , can we apply encoder to categorical variables or is it like only we can apply to a regressor model?, also if we have only few features , is it feasible to apply encoder in classifier models?
@sardarabdullahkhawar84873 жыл бұрын
Sir my question is in categorical variables if we have to use LabelEncoder and there are bad columns like in the exercise then how can we use it in pipeline?? Can you do it in your video part??
@katamit3 жыл бұрын
Hello , how to make column selection step as part of pipe line. i.e.. if I have trained on 10 columns before performing any transformation , i would like to select only those columns. and make this part of pipeline.. Tried using columnTransformer. but columns Transformed return only the numpy array - so the following step of numeric_transformation and categorical _transformation are breaking
@jsklair3 жыл бұрын
Would it have been worth checking the MAE of the GradientBoostingRegressor model on it's own, to see whether it was lower than the MAE of the average of the two models? Thanks for all the videos.
@abhishekkrthakur3 жыл бұрын
we did it on day 10 or 11. but yeah
@kiransethi51093 жыл бұрын
Hii sir , what is the significance of making avearge of the predictions ?
@abhishekkrthakur3 жыл бұрын
in each fold we train on 80% of training data. so, we miss the information from the other 20%. we compensate for this by taking average of predictions from all folds. if you have a lot of data, this might not be needed. this is also a kind of blending, in which you train on subsets of data and then take a simple average of all resulting models (see competition part 5 video)
@Orchishman3 жыл бұрын
In of your previous videos we created a custom imputer where we made the model predict the null values...if we are using Pipelines, is it possible to use such custom functions as imputers?
@abhishekkrthakur3 жыл бұрын
good question! yes you can do that. you need to create a class with "fit" and "transform" functions. Unfortunately, its an advanced concept and out of scope of this series. check out: stackoverflow.com/questions/43232506/using-pipeline-with-custom-classes-in-sklearn
@Orchishman3 жыл бұрын
@@abhishekkrthakur thanks for that prompt response!
@nischaypatel48 ай бұрын
Can you please share the solution of the exercise of Missing values on how to predict the null values by building a model on rest of the features??