Multioutput regression with MLPRegressor - Does it work?

Question

I apologize if this has been addressed before, but I did not find a solution for this issue.

I am currently trying to solve a problem with four dependent variables and four independent variables, all non-categorial. They seem to have non-linear dependence and multicolinearity - standard multivariate regression yielded reasonable results, but I am trying to compare them with results from Random Forest and Neural Networks.

According to this link, it should be possible to get multi-output with the MLPRegressor. However, when I try to apply the .fit method, I get an error message:

fit() missing 1 required positional argument: 'y'

Here is the code that I am using:

# Data Preprocessing - Importing the Libraries
import pandas as pd

# Importing the dataset
dataset = pd.read_csv('Rock_Properties_4T.csv')

y = dataset.iloc[:, 4:8].values
X = dataset.iloc[:, 0:4].values

# ANN
from sklearn.neural_network import MLPRegressor
regressor_ANN = MLPRegressor(tol = 1e-16)
MLPRegressor.fit(X, y)

There you go... Any idea of what could be the problem?

Answer 1

Not sure if this makes a difference, but I think you need to call fit from the initialized object. So for example, your fit line should be:

regressor_ANN.fit(X,y) instead of MLPRegressor.fit(X, y)

then predicting a new set will be:

regressor_ANN.predict(X_test)

Answer 2

I've never encountered that but am guessing from how the documentation is phrased later on with a single target (not targets), you might need to specifically use the multioutput regressor in sklearn like here:

http://scikit-learn.org/stable/auto_examples/ensemble/plot_random_forest_regression_multioutput.html#sphx-glr-auto-examples-ensemble-plot-random-forest-regression-multioutput-py