Improvements to train function and predict function

[dongsik2026]

There are suggestions in the process of using the split data function to distinguish training data from test data and to use training functions and predictive functions.

When creating a train function, it is very convenient to take train_table data of split data by default. In the predict function, however, it is initialized as train_table data, so it needs to be changed to test data every time.

Therefore, it seems to be much easier to set the initial value of test_table to default value in the predict function.