eval_classifier

In this module, methods are defined for evluating TwinSVM perfomance such as cross validation train/test split, grid search and generating the detailed result.

Functions

eval_metrics(y_true, y_pred)	Input:
grid_search(search_space, func_validator)	It applies grid search which finds C and gamma paramters for obtaining best classification accuracy.
initializer(user_input_obj)	It gets user input and passes function and classes arguments to run the program Input: user_input_obj: User input (UserInput class)
save_result(file_name, validator_obj, …)	It saves detailed result in spreadsheet file(Excel).
search_space(kernel_type, class_type, …[, …])	It generates combination of search elements for grid search Input: kernel_type: kernel function which is either linear or RBF c_l_bound, c_u_bound: Range of C penalty parameter for grid search(e.g 2^-5 to 2^+5) rbf_lbound, rbf_ubound: Range of gamma parameter Output: return search elements for grid search (List)

Classes

Validator(X_train, y_train, validator_type, …)

It applies a test method such as cross validation on a classifier like TSVM

eval_classifier.eval_metrics(y_true, y_pred)

Input:

y_true: True label of samples y_pred: Prediction of classifier for test samples

output: Elements of confusion matrix and Evalaution metrics such as accuracy, precision, recall and F1 score

class eval_classifier.Validator(X_train, y_train, validator_type, obj_tsvm)

Bases: object

It applies a test method such as cross validation on a classifier like TSVM

cv_validator(dict_param)

It applies cross validation on instance of Binary TSVM classifier Input:

dict_param: A dictionary of hyper-parameters (dict)

output:

Evaluation metrics such as accuracy, precision, recall and F1 score for each class.

split_tt_validator(dict_param)

It trains TwinSVM classifier on random training set and tests the classifier on test set. output:

Evaluation metrics such as accuracy, precision, recall and F1 score for each class.

cv_validator_mc(dict_param)

It applies cross validation on instance of multiclass TSVM classifier

choose_validator()

It returns choosen validator method.

eval_classifier.search_space(kernel_type, class_type, c_l_bound, c_u_bound, rbf_lbound, rbf_ubound, step=1)

It generates combination of search elements for grid search Input:

kernel_type: kernel function which is either linear or RBF c_l_bound, c_u_bound: Range of C penalty parameter for grid search(e.g 2^-5 to 2^+5) rbf_lbound, rbf_ubound: Range of gamma parameter

Output:

return search elements for grid search (List)

eval_classifier.grid_search(search_space, func_validator)

It applies grid search which finds C and gamma paramters for obtaining best classification accuracy.

Input:

search_space: search_elements (List) func_validator: Validator function

output:

returns classification result (List)

eval_classifier.save_result(file_name, validator_obj, gs_result, output_path)

It saves detailed result in spreadsheet file(Excel).

Input:

file_name: Name of spreadsheet file col_names: Column names for spreadsheet file gs_result = result produced by grid search output_path: Path to store the spreadsheet file.

output:

returns path of spreadsheet file

eval_classifier.initializer(user_input_obj)

It gets user input and passes function and classes arguments to run the program Input:

user_input_obj: User input (UserInput class)