Why there is no need to reshaping of 1D array to fit in linear regression with multiple variables

Question

In sklearn, in order to train the data using fit method in linear regression, we have to reshape the 1D arrays. But, in the case of linear regression with multiple variables, I got the output without reshaping of the target variable.

import numpy as np
import pandas as pd
from sklearn.linear_model import LinearRegression

df = pd.read_csv("Weather.csv",low_memory = False) # the data set I used
df1 = df[["Precip","MaxTemp"]]

reg = LinearRegression().fit(df1.head(),df.MinTemp.head()) # no error with shape of df1 is (5,2) and shape of df.MinTemp.head() is (5,)

Can I know the reason behind this? Thanks in advance.

Fit is taking df1.head() which is X (features) and df.MinTemp.head() as Y(value) ,can you show us where exactly error occurs please have a look and show your error https://stackoverflow.com/questions/45704226/what-does-fit-method-in-scikit-learn-do — nithin, Nov 15 '19 at 10:23

score 2 · Answer 1 · edited Aug 30 '20 at 18:14

please look at this example

from sklearn.linear_model import LinearRegression
import numpy as np

Xr = np.random.randint(0,10,4) # random 1D array with 0 columns -> shape is (4,)

X = np.array([[1, 1], [1, 2], [2, 2], [2, 3]]) # random array which shape is (4,2)
y = np.dot(X, np.array([1, 2])) + 3 # 1D array  with 0 columns -> shape is (4,)

regr = LinearRegression()
# regr.fit(Xr,y) this would raise an exception
regr.fit(X,y)
regr.predict(X) # returns array([ 6.,  8.,  9., 11.])

if you try to fit model using X and y.. you will not get error even if y shape is (4,).it is because your X has the shape of (4,2).so sklearn will cast your targets to X’s dtype if necessary.

if you try to fit Xr and y.both are 1D arrays with 0 columns. You will get error saying Expected 2D array, got 1D array instead:.In this case at least your Xr training data should be a 1D array with 1 column.Then sklearn will do the rest for you by casting target looking at Xr shape.

read more about fit method and what it does here

Why there is no need to reshaping of 1D array to fit in linear regression with multiple variables

1 Answers1