def update_res(config_results, mydir, latest): if (os.path.exists(latest)): shutil.rmtree(latest) text = 'var configs = [' for config in config_results[:-1]: text += str(config) + ',' text += str(config_results[-1]) + '];' if (os.path.exists('config.js')): os.remove('config.js') Metrics.saveConfig('config.js', text) Metrics.copyDirectory(mydir, latest)
#MSE (Training and Validation) Metrics.plot_mse_curve(np.array(error_train), np.array(error_valid), configDir) #Area Under ROC Curve roc_area = Metrics.plot_roc_curve(targetByClass, prob_predictions, configDir) #precision acurracy = ((len(base['testing']['data'])-errors_total)/len(base['testing']['data']))*100 print("acurracy:", acurracy,'%') print('errors',errors_total,'of', len(base['testing']['data'])) configDesc = {'opt_samp':opt_samp.name, 'opt_learning':opt_learning, 'activation_function_options':opt_actvfunc, 'topology_options':opt_top} current_config_result = {'config':configDesc, 'results':{'mse':test_mse,'confusion':{'true_positive':confusion_matrix_percentage[0][0],'false_positive':confusion_matrix_percentage[0][1],'false_negative':confusion_matrix_percentage[1][0],'true_negative':confusion_matrix_percentage[1][1]},'roc':roc_area,'precision':acurracy}} config_results.append(current_config_result.copy()) Metrics.saveConfig(os.path.join(configDir, 'config-results.json'), current_config_result) nConfig = nConfig+1 current_config_result = {} #reset databases from sampling changes training = training_bkp.copy() validation = validation_bkp.copy() testing = testing_bkp.copy() update_res(config_results, mydir, latest) update_res(config_results, mydir, latest)