Training and test split for time series analysis

Question

I need help with below code:

dataset.tail(4)

                Date    Close
    1274    1/24/2019   4871.96
    1275    1/25/2019   4925.82
    1276    1/28/2019   4888.58
    1277    1/29/2019   4928.18

I want to split my dataset to training and test using:

training_data=dataset[dataset.Date<datetime(2018, 1, 1)]

But I get the following error:

TypeError                                 
Traceback (most recent call last)
<ipython-input-74-904bd8b45617> in <module>()
----> 1 training_data=dataset[dataset.Date<datetime(2018, 1, 1)]
      2 training_data.tail(4)

TypeError: 'module' object is not callable

Answer 1

Your code looks incomplete but you can definitely try the following to split your dataset:

X_train, X_test, y_train, y_test = train_test_split(dataset, 
                                                    y,
                                                    test_size=0.3, 
                                                    shuffle=False)

Note: y will be a series object for your dependent variable.

Because of this reason, using train_test_split with shuffle=True is not a good practice and should strictly be avoided.

Alternately, for time series you can use TimeSeriesSplit() and look for the implementation here.

Answer 2

It appears that you're importing the datetime module, then using that to compare dates. The module itself cannot be applied, rather, you must access one of its functions such as:

training_data=dataset[dataset.Date<datetime.datetime(2018, 1, 1)]

An alternative would be:

training_data=dataset[pd.to_datetime('2018-01-01')]

Answer 3

You can keep your code as is and just change the import:

from datetime import datetime
training_data=dataset[dataset.Date<datetime(2018, 1, 1)]