Python: Predicting vector from a scalar-CodePudding

I am working on a problem where I have to predict a vector y from a scalar x. I am currently using linear regression to create a baseline model. But it does not seem to handle the multi-dimesional output.

I am using the following the code:

from sklearn.linear_model import LinearRegression
lm = LinearRegression()

lm.fit(x_train, y_train)

In this case, x_train is a column vector of shape (1,m) and y_train is a vector of vectors of shape (m,).

The error message produced can be seen here.

I think it has to do something with multi-output parameter. Is there any way to work around this?

CodePudding user response：

I am not sure about the data that you are using but try reshaping your "x_train" by using x_train.reshape(-1,1) and then fitting this reshaped data to your linear regression.

CodePudding user response：

Not sure what your underlying data look like, but the error message bool object has no attribute 'any', suggests something wrong with the way your y object is formatted. Here's a minimal working example:

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

N = 100
x = np.random.choice(range(3), size=N)
df = pd.DataFrame({
    'x': x,
    'y1': 3 * x   np.random.normal(size=N),
    'y2': -0.2 * x   np.random.normal(scale=0.2, size=N),
    'y3': -x   12   np.random.normal(scale=0.3, size=N)})
print(df.head())

X = df.pop('x').to_numpy().reshape(-1, 1)
reg = LinearRegression().fit(X, df)

preds = reg.predict(X)
print(preds[:5])
print('coefs:', reg.coef_)
print('intercepts:', reg.intercept_)