Predicting bulldozer price - Wrong Hyperparameters Tuning

[DarkDk123]

In lecture no. 196, you use RandomizedSearchCV with default cv=5 for tuning hyperparameters, i think that's a wrong approach for time series data! because :

• It will perform cross validation by randomly splitting the data into 5-folds i.e. losing intrinsic order of data
• This will result in poor evaluation of best hyperparameters

What ChatGPT says -

What we can do is use a TimeSeriesSplit of sklearn!

You should suggest the correct way of doing this in you course soon!