SVR_class document:

Author Info:

• Name: Kent010341 (Nickname using on internet)
• e-mail: [email protected]

---

Parameters:

• kernel: string or callable method, optional (default='rbf')
  • Kernel type, currently support linear, rbf kernel and custom kernel.
  • Custom kernel must satisfy the Mercer's conditions.
• C: float, optional (default=1)
  • Penalty parameter C of the error term.
• gamma: float, optional (default='auto')
  • Coefficient of kernel like rbf kernel.
• epsilon: float, optional (default=0.1)
  • Epsilon in epsilon-SVR
• max_iter: int, optional (default=-1)
  • Maximum iterations to solving problem, using -1 for no limit.
• debug: bool, optional (default=False)
  • Print all traininig detail if debug is True.
• random_seed: int, optional (default=0)
  • Seed for numpy.random.seed and random.seed.

Attributes:

• kerneled_matrix: 2-D array
  • Training set after kernel transfer.
• alphas: array
  • Coefficients of the support vector.
• b: float
  • Bias of the prediction function.

Methods:

• get_params(self):
  • Print the parameters set by initially defined.
• fit(self, train_X, train_y):
  • Fit the model according to the given training data.
• predict(self, test_X):
  • Perform regression on testing input.
  • Can only be used if the model is fit or loaded.
• save_model(self, save_dir='default', file_name='default'):
  • Save the trained model as json form. Defaultly using "trained_models" as save_dir and time form as file_name.
• load_model(self, file_dir):
  • Load the trained model.

---

Reference:

• Understanding Support Vector Machine Regression
• Efficient SVM Regression Training with SMO
• sklearn.svm.SVR
• SVM code