특성 X와 대상 y

Supervised models learn a mapping from inputs to an output. The inputs are your features, and the output is what you want to predict. 🎯

By convention, features go in a capital X and the target goes in a lowercase y. This naming shows up in nearly every example you read.

Capital X is a 2D table: one row per sample, one column per feature. scikit-learn always expects this rows-by-columns shape.

The target y is usually a single column: one value per sample. Its length must match the number of rows in X exactly.

From a pandas table, you carve out X and y by selecting columns. The target column becomes y; everything else becomes X.

Most estimators expect numeric features. Text and category columns need encoding into numbers before they can enter X.

Never leave the answer inside your features. A target hiding in X is leakage, and it makes a model look perfect but useless.

Row one of X must correspond to value one of y. This alignment is how the model connects each sample to its correct answer.

If y holds continuous numbers it is regression; if y holds categories it is classification. The target's type decides the model family.

Once X and y are ready, you hand both to fit together. The model reads features and answers side by side to learn.

When predicting, you pass new features with the same columns as training X. You never pass y, since that is what you want back.

Make sure you have the shapes of X and y straight.

Split your data into X, a 2D feature table, and y, the 1D target. Keep them aligned and leak-free, and fit is ready to learn. ✅