Simple train test splitting in pandas

split_train_test.py

import pandas as pd
import numpy as np

def split_train_test(df_in, test_ratio):
    # Reserve an out-of-place sample for final evaluation
    # Could have used train_test_split from sklearn
    df = df_in.copy(deep=True)
    np.random.seed(42)
    shuffled_indices = np.random.permutation(len(df))
    test_set_size = int(len(df) * test_ratio)
    test_indices = shuffled_indices[:test_set_size]
    train_indices = shuffled_indices[test_set_size:]
    return df.iloc[train_indices], df.iloc[test_indices]