def modelValid(x_data,y_data): error = 0 err_list =[] x_data = np.reshape(x_data, (-1, 1, 6)) y_list = np.unique(y_data) y_count =[] for j in range(len(y_list)): err_list.append(0) y_count.append(0) model_dir = "Model/Regression/"+load_date+"/" name = model_dir + gas_type + "_reg" cnn = load_model(name+'.pkl') for i in range(len(x_data)): result = cnn.test(torch.unsqueeze(torch.Tensor(x_data[i]).float(), dim=0)).tolist() result_int = int(result[0][0]) err = round(abs(y_data[i] - result_int) / y_data[i] * 100, 1) error = error + err for j in range(len(y_list)): if y_data[i] == y_list[j]: err_list[j] = err_list[j]+err y_count[j] = y_count[j] + 1 print(result_int,"-",y_data[i],":",err) # μΈνΈ for j in range(len(y_list)): print(y_list[j], ":", round(err_list[j] / y_count[j], 1)) print(gas_type, "Error :", round(sum(err_list) / sum(y_count), 1)) print("")
def modelValid(x_data,y_data,name): dnn = load_model(name+'.pkl') dnn.train(False) for i in range(len(x_data)): x = torch.Tensor(x_data[i]).float() x = x.unsqueeze(dim=0) result = dnn.forward(x.cuda()) print(np.round(result.tolist(), 2), y_data[i])
def modelValid(x_data,y_data): error = 0 err_list =[] y_list = np.unique(y_data) y_count =[] for j in range(len(y_list)): err_list.append(0) y_count.append(0) model_dir = "Model/Regression/"+load_date+"/" name = model_dir + gas_type + "_reg" dnn = load_model(name+'.pkl') dnn.train(False) for i in range(len(x_data)): x = torch.Tensor(x_data[i]).float() x = x.unsqueeze(dim=0) result = dnn.forward(x.cuda()) result_int = int(result.tolist()[0][0]) err = round(abs(y_data[i] - result_int)/y_data[i] * 100,1) error = error + err for j in range(len(y_list)): if y_data[i] == y_list[j]: err_list[j] = err_list[j]+err y_count[j] = y_count[j] + 1 print(result_int, "-", y_data[i], ":", err) # μΈνΈ for j in range(len(y_list)): print(y_list[j], ":", round(err_list[j] / y_count[j], 1)) print(gas_type, "Error :", round(sum(err_list) / sum(y_count), 1)) print("")
from datetime import datetime
print("- Port :", Port) print("- DB :", Database) print("- GPU Use :", gpu) print("") class_name = ['Normal','H2S','NH3','CH3SH','CO','CH4','H2S+NH3','H2S+CH3SH','H2S+CO','H2S+CH4','NH3+CH3SH','NH3+CO','NH3+CH4','CH3SH+CO','CH3SH+CH4','CO+CH4'] clf_name = ['model/Classify/Clf_all.pkl', 'model/Classify/Clf_single.pkl', 'model/Classify/Clf_mix.pkl'] reg_name = ['model/Regression/H2S_Reg.pkl', 'model/Regression/NH3_Reg.pkl', 'model/Regression/CH3SH_Reg.pkl', 'model/Regression/CO_Reg.pkl', 'model/Regression/CH4_Reg.pkl'] scaler_model = [joblib.load('model/Classify/all_scaler.pkl'),joblib.load('model/Classify/single_scaler.pkl'),joblib.load('model/Classify/mix_scaler.pkl')] print("- Model Loaded") clf_model = [] reg_model = [] if gpu: for i in range(len(clf_name)): tmp_clf = load_model(clf_name[i]) tmp_clf.train(False) clf_model.append(tmp_clf) print("- DNN CLF Model Loaded") for i in range(len(reg_name)): tmp_reg = load_model(reg_name[i]) tmp_reg.train(False) reg_model.append(tmp_reg) print("- DNN REG Model Loaded") else: for i in range(len(clf_name)): tmp_clf = load_model(clf_name[i], map_location='cpu') tmp_clf.train(False) clf_model.append(tmp_clf) print("- DNN CLF Model Loaded") for i in range(len(reg_name)):