python机器学习 在lstm模型的基础上加入了卷积 可以处理多变量输入的问题

#!/usr/bin/env python # coding: utf-8 # In[ ]: # 查看当前挂载的数据集目录, 该目录下的变更重启环境后会自动还原 # View dataset directory. # This directory will be recovered automatically after resetting environment. get_ipython().system('ls /home/aistudio/data') # In[ ]: # 查看工作区文件, 该目录下的变更将会持久保存. 请及时清理不必要的文件, 避免加载过慢. # View personal work directory. # All changes under this directory will be kept even after reset. # Please clean unnecessary files in time to speed up environment loading. get_ipython().system('ls /home/aistudio/work') # In[ ]: # 如果需要进行持久化安装, 需要使用持久化路径, 如下方代码示例: # If a persistence installation is required, # you need to use the persistence path as the following: get_ipython().system('mkdir /home/aistudio/external-libraries') get_ipython().system('pip install beautifulsoup4 -t /home/aistudio/external-libraries') # In[1]: # 同时添加如下代码, 这样每次环境(kernel)启动的时候只要运行下方代码即可: # Also add the following code, # so that every time the environment (kernel) starts, # just run the following code: import sys sys.path.append('/home/aistudio/external-libraries') import numpy as np import pandas as pd import matplotlib.pyplot as plt import paddle import paddle.nn as nn from sklearn.preprocessing import MinMaxScaler # 标签就是车流量traffic_volume 特征是is_holiday is_weekend temp clouds_all df = pd.read_csv('./dataset/series1.csv', usecols=['traffic_volume','temp', 'holiday', 'is_weekend']) series_test = pd.read_csv('./dataset/series_test.csv', usecols=['temp', 'holiday', 'is_weekend']).values all_data = df.values split_boundary = int(all_data.shape[0] * 0.98) train_data = all_data[: split_boundary, :] test_data = all_data[split_boundary:, :] plt.figure(figsize=(15, 8)) # 数据可视化 # plt.plot(np.arange(train_data.shape[0]), train_data[:, 0], label='train data') # plt.plot(np.arange(train_data.shape[0], train_data.shape[0] + test_data.shape[0]), test_data[:, 0], label='test data') # plt.legend() # 窗口划分 def split_windows(x1,y1, size): X = [] Y = [] # X作为数据，Y作为标签 # 滑动窗口，步长为1 for i in range(len(x1) - size): X.append(x1[i:i + size, :]) Y.append(y1[i + size, 0]) return np.array(X), np.array(Y) def split_test(data, size): X = [] # X作为数据 # 滑动窗口，步长为1 for i in range(len(data) - size): X.append(data[i:i + size, :]) return np.array(X) def fit_size(x, y): from sklearn import preprocessing x_MinMax = preprocessing.MinMaxScaler() y_MinMax = preprocessing.MinMaxScaler() x = x_MinMax.fit_transform(x) y = y_MinMax.fit_transform(y) return x, y, y_MinMax # normalizatioin processing train_x = train_data[:,1:] train_y = train_data[:,0].reshape(-1, 1) test_x = test_data[:,1:] test_y = test_data[:,0].reshape(-1, 1) train_x, train_y, y_MinMax = fit_size(train_x, train_y) test_x, test_y, y_MinMax1 = fit_size(test_x, test_y) # scaled_train_data = scaler.fit_transform(train_data) # 使用训练集的最值对测试集归一化，保证训练集和测试集的分布一致性 # scaled_test_data = scaler.transform(test_data) scaler = MinMaxScaler() series_test = scaler.fit_transform(series_test) # 训练集测试集划分 window_size = 30 train_X, train_Y = split_windows(train_x,train_y, size=window_size) test_X, test_Y = split_windows(test_x,test_y, size=window_size) series_test = split_test(series_test, size=window_size) print('train shape', train_X.shape, train_Y.shape) print('test shape', test_X.shape, test_Y.shape) print('series_test shape', series_test.shape) # 预测未来一周每小时的数据 def process(data, bs): l = len(data) tmp = [] for i in range(0, l, bs): if i + bs > l: tmp.append(data[i:].tolist()) else: tmp.append(data[i:i + bs].tolist()) tmp = np.array(tmp) return tmp batch_size = 1024 train_X = process(train_X, batch_size) train_Y = process(train_Y, batch_size) print(train_X.shape, train_Y.shape) fea_num = 4 out_fea = 30 class CNN_LSTM(nn.Layer): def __init__(self, window_size, fea_num): super().__init__() self.window_size = window_size self.fea_num = fea_num self.proj = nn.Linear(in_features=fea_num, out_features=out_fea) self.conv1 = nn.Conv2D(in_channels=1, out_channels=64, stride=1, kernel_size=3, padding='same') self.relu1 = nn.ReLU() self.pool = nn.MaxPool2D(kernel_size=2, stride=1, padding='same') self.dropout = nn.Dropout2D(0.3) self.lstm1 = nn.LSTM(input_size=64 * out_fea, hidden_size=256, num_layers=1, time_major=False) self.lstm2 = nn.LSTM(input_size=256, hidden_size=128, num_layers=1, time_major=False) self.lstm3 = nn.LSTM(input_size=128, hidden_size=64, num_layers=1, time_major=False) self.fc = nn.Linear(in_features=64, out_features=32) self.relu2 = nn.ReLU() self.head = nn.Linear(in_features=32, out_features=1) def forward(self, x): x = x.reshape([x.shape[0], 1, self.window_size, self.fea_num]) x = self.proj(x) x = self.conv1(x) x = self.relu1(x) x = self.pool(x) x = self.dropout(x) x = x.reshape([x.shape[0], self.window_size, -1]) x, (h, c) = self.lstm1(x) x, (h, c) = self.lstm2(x) x, (h, c) = self.lstm3(x) x = x[:, -1, :] x = self.fc(x) x = self.relu2(x) x = self.head(x) return x model = CNN_LSTM(window_size, fea_num) # paddle.summary(model, (99, 30, 4)) # 定义超参数 base_lr = 0.001 EPOCH = 100 lr_schedual = paddle.optimizer.lr.CosineAnnealingDecay(learning_rate=base_lr, T_max=EPOCH, verbose=True) loss_fn = nn.MSELoss() mae_loss = nn.L1Loss() metric = paddle.metric.Accuracy() # opt = paddle.optimizer.SGD(parameters=model.parameters(),learning_rate=lr_schedual) opt = paddle.optimizer.Adam(parameters=model.parameters(), learning_rate=lr_schedual, beta1=0.9, beta2=0.999) mses_train= [] # mses_eval= [] maes_train= [] # maes_eval= [] # 模型训练 for epoch in range(EPOCH): model.train() loss_train = 0 for batch_id, data in enumerate(train_X): label = train_Y[batch_id] data = paddle.to_tensor(data, dtype='float32') label = paddle.to_tensor(label, dtype='float32') label = label.reshape([label.shape[0], 1]) y = model(data) loss = loss_fn(y, label) opt.clear_grad() loss.backward() opt.step() loss_train += loss.item() mae = mae_loss(y,label) mses_train.append(loss.item()) maes_train.append(mae.item()) print("[TRAIN] ========epoch : {}, loss: {:.4f}==========".format(epoch + 1, loss_train)) lr_schedual.step() # loss_eval = 0 # # model.eval() # for batch_id, data in enumerate(test_X): # label = test_Y[batch_id] # data = paddle.to_tensor(data, dtype='float32') # label = paddle.to_tensor(label, dtype='float32') # label = label.reshape([label.shape[0],1]) # y = model(data) # mae = mae_loss(y,label) # loss = loss_fn(y, label) # loss_eval += loss.item() # mses_eval.append(loss.item()) # maes_eval.append(mae.item()) # print("[EVAL] ========epoch : {}, loss: {:.4f}==========\n".format(epoch+1, loss_eval)) # 保存模型参数 paddle.save(model.state_dict(), 'wind/cnn_lstm1_ep{}_lr{}.params'.format(EPOCH,base_lr)) paddle.save(lr_schedual.state_dict(), 'wind/cnn_lstm1_ep{}_lr{}.pdopts'.format(EPOCH,base_lr)) # 加载模型 model = CNN_LSTM(window_size, fea_num) model_dict = paddle.load('wind/cnn_lstm1_ep{}_lr{}.params'.format(EPOCH,base_lr)) model.load_dict(model_dict) test_X = paddle.to_tensor(test_X, dtype='float32') series_test =