Ensemble multiple models using VotingClassifer or VotingRegressor

Рет қаралды 10,137

Күн бұрын

Пікірлер: 16

@dataschool 2 жыл бұрын

Thanks for watching! 🙌 To be clear, ensembling can improve your model regardless of whether your chosen evaluation metric is accuracy or a different metric. In other words, it's always worth a try!

@grzegorzzawadzki8718 2 жыл бұрын

I was looking for solutions on how to improve the quality and readability of code using pipelines in my project and I came across your channel. It's a shame that you didn't show how to apply these methods on larger datasets, but I still find your videos very useful. The thing I'm struggling the most is how to apply cleaning and features engineering and creating new features using pipelines.

@Prajwal_KV 2 жыл бұрын

Learned a lot.Thanks

@dataschool 2 жыл бұрын

Glad to hear it!

@y4173 2 жыл бұрын

Loved it!

@dataschool 2 жыл бұрын

Thank you!

@shakilahmad2640 Жыл бұрын

hi, Sir can you please tell me how we can ensemble VGG16 and MOBILENET ?

@dataschool Жыл бұрын

I'm not familiar with those, I'm sorry!

@elouassifbouchra5079 Жыл бұрын

What is the minimum number of base learners in an ensemble three or two? Thanks in advance

@KevinMarkham Жыл бұрын

Two!

@aleksandartta 2 жыл бұрын

Thank you... I understand what is random_state (you explained that is important for reproducibility). But, is it necessary here? It make sense when we split the data, but why is here important? If I am not wrong these models without the random_state would be always the same? Thank you in advance

@dataschool 2 жыл бұрын

Both LogisticRegression (when using the liblinear solver) and RandomForestClassifier have randomness in the model building. Hope that helps!

@21Gannu 2 жыл бұрын

real OG of data science

@dataschool 2 жыл бұрын

You are too kind!

@attilasarkany6123 Жыл бұрын

Can we use voting regression with time series data?

@dataschool Жыл бұрын

You can use ensembling regardless of the type of data. That being said, a typical supervised ML model in scikit-learn may not be the best fit for time series data, depending on your goals.