How should the data scientist split the dataset into a training dataset and a validation dataset to compare model performance?
Pick a date so that 80% of the data points precede the date. Assign that group of data points as the training dataset. Assign all the remaining data points to the validation dataset.
Pick a date so that 80% of the data points occur after the date. Assign that group of data points as the training dataset. Assign all the remaining data points to the validation dataset.
Starting from the earliest date in the dataset, pick eight data points for the training dataset and two data points for the validation dataset. Repeat this stratified sampling until no data points remain.
Sample data points randomly without replacement so that 80% of the data points are in the training dataset. Assign all the remaining data points to the validation dataset.
Explanations:
This method preserves the time series order by assigning the earlier 80% of the data to the training set and the later 20% to the validation set. This is critical for forecasting models, as future data cannot be used to predict past data.
This method assigns the later 80% of the data to the training set, which violates the principle of time series forecasting. Using future data to predict past data leads to data leakage and unrealistic model performance.
This method uses stratified sampling, which is not suitable for time series data. In time series forecasting, the order of data points matters, and random sampling can disrupt the temporal sequence, leading to unrealistic predictions.
Random sampling without replacement is not appropriate for time series data, as it disrupts the chronological order. Forecasting models require training on past data to predict future data, making this approach invalid for the given task.
I map out that the answer is:
Pick a date so that 80% of the data points precede the date. Assign that group of data points as the training dataset. Assign all the remaining data points to the validation dataset.