sin信号输入输出.rar

import torch import torch.nn as nn import torch.optim as optim import numpy as np import matplotlib.pyplot as plt import pandas as pd from sklearn.preprocessing import MinMaxScaler from sklearn.metrics import mean_squared_error, mean_absolute_error import math # 输入数据时序步数 STEP = 4 # gauss状态分布随机信号数部分(origin) data = pd.read_csv('12-06 15_54_07_sibato.csv') # 读取文件 training_set = data.iloc[200:5212 - 1200, 1:3].values training_set[:, 0] = training_set[:, 0] - 2.305 test_set_org = data.iloc[5211 - 1200:5212 - 500, 1:3].values test_set_org[:, 0] = test_set_org[:, 0] - 2.305 ''' # gauss状态分布随机信号数部分(my_measurement) data = pd.read_csv('data_test.csv') # 读取文件 training_set = data.iloc[200:3900, 1:3].values test_set_org = data.iloc[3900:4300, 1:3].values ''' # sin信号部分 # data1 = pd.read_csv('data_sin_test1.csv') # training_set = data1.iloc[:1760, 1:3].values # test_set_org = data1.iloc[1760:, 1:3].values # sc = MinMaxScaler(feature_range=(0, 1)) # training_set_scaled1 = sc.fit_transform(training_set) # test_set = sc.transform(test_set_org) # pre_set = sc.transform(test_set_org) # -------------------------------------------------------------- sc = MinMaxScaler(feature_range=(0, 1)) training_set_scaled = sc.fit_transform(training_set) test_set = sc.transform(test_set_org) pre_set = sc.transform(test_set_org) x_train = [] y_train = [] x_test = [] y_test = [] for i in range(STEP, len(training_set_scaled)): x_train.append(training_set_scaled[i - STEP:i, 0:2]) y_train.append(training_set_scaled[i, 0]) x_train = np.array(x_train) y_train = np.array(y_train) x_train = torch.tensor(x_train, dtype=torch.float32) y_train = torch.tensor(y_train, dtype=torch.float32) x_test = [] y_test = [] for i in range(STEP, len(test_set)): x_test.append(test_set[i - STEP:i, 0:2]) y_test.append(test_set[i, 0]) x_test = np.array(x_test) y_test = np.array(y_test) x_test = torch.tensor(x_test, dtype=torch.float32) class SimpleRNNModel(nn.Module): def __init__(self): super(SimpleRNNModel, self).__init__() self.rnn1 = nn.LSTM(input_size=2, hidden_size=5, batch_first=True, num_layers=2, dropout=0.2)#RNN self.rnn2 = nn.LSTM(input_size=5, hidden_size=10, batch_first=True, num_layers=2, dropout=0.2)#RNN self.linear = nn.Linear(10, 1) def forward(self, x): out, _ = self.rnn1(x) out, _ = self.rnn2(out) out = self.linear(out[:, -1, :]) # 取每个序列最后一个时间步的输出 return out model = SimpleRNNModel() criterion = nn.MSELoss() optimizer = optim.Adamax(model.parameters(), lr=0.002) # Adamax Adagrad ASGD Adadelta num_epochs = 1500 for epoch in range(num_epochs): optimizer.zero_grad() outputs = model(x_train) loss = criterion(outputs, y_train.view(-1, 1)) # 需要将 y_train 变形成 [batch_size, 1] 形状 #loss = criterion(outputs, y_train) # 需要将 y_train 变形成 [batch_size, 1] 形状 loss.backward() optimizer.step() print(f"Epoch [{epoch + 1}/{num_epochs}], Loss: {loss.item():.6f}") # 测试集输入模型进行预测 with torch.no_grad(): predicted_force = model(x_test) pre_set[STEP:, 0] = predicted_force[:, 0] predicted = pre_set predicted_force = sc.inverse_transform(predicted) predicted_force = predicted_force[STEP:, 0] real_force = sc.inverse_transform(test_set[STEP:, :]) real_pwm = real_force[:, 1] real_force = real_force[:, 0] mse = mean_squared_error(predicted_force, real_force) rmse = math.sqrt(mean_squared_error(predicted_force, real_force)) mae = mean_absolute_error(predicted_force, real_force) print('均方误差: %.6f' % mse) print('均方根误差: %.6f' % rmse) print('平均绝对误差: %.6f' % mae) # 画出真实数据和预测数据的对比曲线 plt.plot(-real_force[:, ], color='red', label='Actually Force') plt.plot(-predicted_force[:, ], color='blue', label='Predicted Force') plt.plot(-real_pwm[:, ], color='green', label='Actually Input') plt.grid plt.xlabel('Time') plt.ylabel('Force') plt.legend() plt.show() ''' # 测试集输入模型进行预测(用sin信号测试） with torch.no_grad(): predicted_force = model(x_test1) pre_set1[STEP:, 0] = predicted_force[:, 0] predicted = pre_set1 predicted_force = sc1.inverse_transform(predicted) predicted_force = predicted_force[STEP:, 0] real_force = sc.inverse_transform(test_set1[STEP:, :]) real_pwm = real_force[:, 1] real_force = real_force[:, 0] mse = mean_squared_error(predicted_force, real_force) rmse = math.sqrt(mean_squared_error(predicted_force, real_force)) mae = mean_absolute_error(predicted_force, real_force) print('均方误差: %.6f' % mse) print('均方根误差: %.6f' % rmse) print('平均绝对误差: %.6f' % mae) # 画出真实数据和预测数据的对比曲线 plt.plot(-real_force[:, ], color='red', label='Actually Force') plt.plot(-predicted_force[:, ], color='blue', label='Predicted Force') plt.title('Force Prediction') plt.xlabel('Time') plt.ylabel('Force') plt.legend() plt.show() '''