TopolearnMost beginner errors in scikit-learn are shape errors; knowing the convention prevents them.
scikit-learn expects data in a fixed layout. X is a 2-D array-like of shape (n_samples, n_features): one row per example, one column per feature. y is a 1-D array of shape (n_samples,) holding the target for each row: a class label for classification, a number for regression.
X can be a NumPy array, a pandas DataFrame or a SciPy sparse matrix. When you pass a DataFrame, the estimator remembers the column names in feature_names_in_ and warns if you later predict on data with different columns. y is usually a NumPy array or a pandas Series.
The most common error is passing a single feature as a 1-D array. model.fit(df["age"], y) fails with "Expected 2D array, got 1D array instead". Keep it 2-D:
df[["age"]](double brackets) gives a one-column DataFrame,x.reshape(-1, 1)turns a 1-D NumPy array into a column.
The reverse also happens: passing y as a one-column DataFrame or (n, 1) array triggers a DataConversionWarning; use df["target"] or y.ravel().
For prediction you always pass a 2-D X too, even for one sample: model.predict([[5.1, 3.5]]), not model.predict([5.1, 3.5]). All features must be numeric unless the estimator explicitly supports categories, which is why encoding is a separate step.
Example
import numpy as np
import pandas as pd
from sklearn.linear_model import LinearRegression
df = pd.DataFrame({"area": [50, 80, 120, 65], "rooms": [2, 3, 4, 2],
"price": [150, 240, 350, 190]})
X = df[["area", "rooms"]] # 2-D: (4, 2)
y = df["price"] # 1-D: (4,)
print(X.shape, y.shape) # (4, 2) (4,)
model = LinearRegression().fit(X, y)
print(model.feature_names_in_) # ['area' 'rooms']
new = pd.DataFrame({"area": [100], "rooms": [3]})
print(model.predict(new).round(1)) # [289.2]
area = np.array([50, 80, 120, 65])
print(area.reshape(-1, 1).shape) # (4, 1) - a single feature, still 2-DRemember
- X is 2-D
(n_samples, n_features), y is 1-D(n_samples,) - Use
df[["col"]]orreshape(-1, 1)for a single feature - Predict on 2-D input even for one row
Watch out for
- Calling
fit(df["col"], y)with a 1-D Series as X.