CODEWITHSUNDEEP
pythonmachine-learningsvmgrid-search
Ganesh
Ganesh

Reputation: 33

Invalid Parameter loss for estimator SVR

This is my code

i used grid search cv for hyper parameter tuning. but it shows error.

param_grid = {"kernel" : ['linear', 'poly', 'rbf', 'sigmoid'],
            'loss' : ['epsilon_insensitive', 'squared_epsilon_insensitive'],
             "max_iter" : [1,10,20],
             'C' : [np.arange(0,20,1)]} 

model = GridSearchCV(estimator = svr, param_grid = param_grid, cv = 5, verbose = 3, n_jobs = -1)

m1 = model.fit(x_train,y_train)

ValueError: Invalid parameter loss for estimator SVR(C=array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16,
   17, 18, 19]),
kernel='linear'). Check the list of available parameters with `estimator.get_params().keys()`.

Upvotes: 0

Views: 573

Answers (1)

yatu
yatu

Reputation: 88295

Some errors that I spotted:

  • You seem to be specifying a loss parameter and possible values, that are only defined for a LinearSVR, not a SVR. On another hand, if you do want to use a LinearSVR, you can't specify a kernel, since it has to be linear.

  • I also noticed that 'C' : [np.arange(0,20,1)] in the definition of the grid will yield an error, since it results in a nested list. Just use np.arange(0,20,1)

Assuming then you have a SVR, the following should work for you:

from sklearn.svm import SVR
svr = SVR()

param_grid = {"kernel" : ['linear', 'poly', 'rbf', 'sigmoid'],
             "max_iter" : [1,10,20],
             'C' : np.arange(0,20,1)} 

model = GridSearchCV(estimator = svr, param_grid = param_grid, 
                     cv = 5, verbose = 3, n_jobs = -1)
m1 = model.fit(X_train, y_train)

Upvotes: 1

Related Questions