Myujth

I want to create a contour plot for a prediction with multiple features. The remaining values should be fixed to plot the 2 interesting values. Unfortunately I resulting matrix has the same value on all positions instead of the expected.

I think something with my matrixes is wrong, but I don't find the error.

[...]
f_learn = [x_1,x_2,x_3,x_4]
r_lear = [r_1]

clf = svm.MLPRegressor(...)
clf.fit(f_learn,r_learn)
[...]

x_1 = np.linspace(1, 100, 100)
x_2 = np.linspace(1, 100, 100)
X_1, X_2 = np.meshgrid(x_1, x_2)

x_3 = np.full( (100,100), 5).ravel()
x_4 = np.full( (100,100), 15).ravel()

predict_matrix = np.vstack([X_1.ravel(), X_2.ravel(), x_3,x_4])
prediction = clf.predict(predict_matrix.T)

prediction_plot = prediction.reshape(X_1.shape)

plt.figure()
 cp = plt.contourf(X_1, X_2, prediction_plot, 10)
 plt.colorbar(cp)
 plt.show()

If I test the matrix line by line by hand I get the right results. However, it doesn't work if I put them together this way.

Edit: made a error copying the code

Example with Data. All answer are 7.5 and not diffrent ;(

import matplotlib.pyplot as plt
import numpy as np
from sklearn import linear_model

f_learn = np.array([[1, 2, 3, 4], [1, 2, 3, 4], [1, 2, 3, 4], [1, 2, 3, 4]])
r_learn = np.array([6,7,8,9])

reg = linear_model.LinearRegression()
reg.fit (f_learn, r_learn)

x_1 = np.linspace(0, 20, 10)
x_2 = np.linspace(0, 20, 10)
X_1, X_2 = np.meshgrid(x_1, x_2)

x_3 = np.full( (10,10), 5).ravel()
x_4 = np.full( (10,10), 2).ravel()

predict_matrix = np.vstack([X_1.ravel(), X_2.ravel(), x_3, x_4])
prediction = reg.predict(predict_matrix.T)

prediction_plot = prediction.reshape(X_1.shape)

plt.figure()
cp = plt.contourf(X_1, X_2, prediction_plot, 10)
plt.colorbar(cp)
plt.show()

Result

In your toy data, there are 4 examples with the same feature values and different label. LinearRegression doeasn't learn anything from it. You can check it by:

>>> reg.coef_
[0. 0. 0. 0.]

Maybe it is also the case in your real data. That the features x_1, x_2 doesn't matter. Check reg.coef_ if there are not too small values for features x_1, x_2.

I changed the toy data and the plot is working.

import matplotlib.pyplot as plt
import numpy as np
from sklearn import linear_model

# f_learn = np.array([[1, 2, 3, 4], [1, 2, 3, 4], [1, 2, 3, 4], [1, 2, 3, 4]])
# r_learn = np.array([6,7,8,9])
f_learn = np.arange(20.).reshape(5, 4)
f_learn += np.random.randn(5, 4)
r_learn = f_learn[:, 0] + 2 * f_learn[:, 1] + 3 * f_learn[:, 2] + 4 * f_learn[:, 3]

reg = linear_model.LinearRegression()
reg.fit(f_learn, r_learn)
print(reg.coef_)

x_1 = np.linspace(0, 20, 10)
x_2 = np.linspace(0, 20, 10)
X_1, X_2 = np.meshgrid(x_1, x_2)

x_3 = np.full( (10,10), 5).ravel()
x_4 = np.full( (10,10), 2).ravel()

predict_matrix = np.vstack([X_1.ravel(), X_2.ravel(), x_3, x_4])
prediction = reg.predict(predict_matrix.T)

prediction_plot = prediction.reshape(X_1.shape)

plt.figure()
cp = plt.contourf(X_1, X_2, prediction_plot, 10)
plt.colorbar(cp)
plt.show()

Try something like this. Some comments in the code

x_1 = np.linspace(1, 100, 100)
x_2 = np.linspace(1, 100, 100)
X_1, X_2 = np.meshgrid(x_1, x_2)

# Why the shape was (1000, 100)?
x_3 = np.full((100, 100), 5).ravel() 
x_4 = np.full((100, 100), 15).ravel()

# you should use X_1.ravel() to make it column vector (it is one feature)
# there was x_3 insted of x_4
predict_matrix = np.vstack([X_1.ravel(), X_2.ravel(), x_3, x_4]) 
prediction = clf.predict(predict_matrix.T)

prediction_plot = prediction.reshape(X_1.shape)

plt.figure()
cp = plt.contourf(X_1, X_2, prediction_plot, 10)
plt.colorbar(cp)
plt.show()

Following code should give you the contour plot you want.

from sklearn.datasets import make_regression

f_learn, r_learn = make_regression(20,4)

reg = linear_model.LinearRegression()
reg.fit (f_learn, r_learn)

x_1 = np.linspace(-2, 2, 10)
x_2 = np.linspace(-2, 2, 10)
X_1, X_2 = np.meshgrid(x_1, x_2)


x_3 = np.full( (10,10), 0.33).ravel()
x_4 = np.full( (10,10), 0.99).ravel()

predict_matrix = np.vstack([X_1.ravel(), X_2.ravel(), x_3, x_4])
prediction = reg.predict(predict_matrix.T)

prediction_plot = prediction.reshape(X_1.shape)

plt.figure()
cp = plt.contourf(X_1, X_2, prediction_plot, 10)
plt.colorbar(cp)
plt.show()