Data Science · Chapter 25 of 43
Train / Test Split
SPLIT your data into TRAIN (fit the model) and TEST (evaluate honestly). Typical ratios: 80/20 or 70/30.
Never peek at the test set while iterating — that inflates your reported accuracy.
Example 1 (python)
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)Reproducible split.
Example 2 (python)
# For time series, split by TIME not randomlyChronological split for time-ordered data.
Key points
- Always split before training.
- 80/20 or 70/30 is typical.
- random_state for reproducibility.
- For time series, split chronologically.
💡 Note: For CLASSIFICATION with imbalanced classes, use `stratify=y` to keep the class ratio in both splits.