7.1 MiB
7.1 MiB
In [1]:
import os os.environ['CUDA_DEVICE_ORDER'] = 'PCB_BUS_ID' os.environ['CUDA_VISIBLE_DEVICES'] = '0'
In [2]:
import pandas as pd import numpy as np from sklearn.model_selection import train_test_split import matplotlib.pyplot as plt #新增加的两行 from pylab import mpl # 设置显示中文字体 mpl.rcParams["font.sans-serif"] = ["SimHei"] mpl.rcParams["axes.unicode_minus"] = False
In [3]:
data = pd.read_csv('./new_train_data.csv', index_col='date') # data.drop(columns=['wd'], inplace=True) # 风向还没想好怎么处理 data.head()
Out[3]:
| 24_PM2.5 | 24_PM10 | 24_SO2 | 24_NO2 | 24_O3 | 24_CO | 23_PM2.5 | 23_PM10 | 23_SO2 | 23_NO2 | ... | NH3_resdient | NH3_agricultural | VOC_industrial | VOC_transportation | VOC_resdient | VOC_power | PM2.5_industrial | PM2.5_transportation | PM2.5_resdient | PM2.5_power | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| date | |||||||||||||||||||||
| 2015-01-03 01:00:00 | 136.0 | 214.0 | 317.0 | 38.0 | 8.0 | 3.71 | 114.0 | 176.0 | 305.0 | 38.0 | ... | 0.033910 | 0.359273 | 1.177423 | 1.084925 | 0.937173 | 0.037724 | 0.926851 | 0.077715 | 0.827110 | 0.436028 |
| 2015-01-03 02:00:00 | 114.0 | 176.0 | 305.0 | 38.0 | 8.0 | 3.55 | 97.0 | 154.0 | 306.0 | 37.0 | ... | 0.033910 | 0.359273 | 1.177423 | 1.134240 | 0.937173 | 0.036215 | 0.926851 | 0.081248 | 0.827110 | 0.418587 |
| 2015-01-03 03:00:00 | 97.0 | 154.0 | 306.0 | 37.0 | 7.0 | 3.51 | 87.0 | 141.0 | 316.0 | 38.0 | ... | 0.033910 | 0.327791 | 1.177423 | 1.232869 | 0.937173 | 0.035712 | 0.926851 | 0.088313 | 0.827110 | 0.412773 |
| 2015-01-03 04:00:00 | 87.0 | 141.0 | 316.0 | 38.0 | 7.0 | 3.55 | 85.0 | 139.0 | 292.0 | 37.0 | ... | 0.033910 | 0.350014 | 1.177423 | 1.273965 | 0.937173 | 0.036718 | 0.926851 | 0.091256 | 0.827110 | 0.424400 |
| 2015-01-03 05:00:00 | 85.0 | 139.0 | 292.0 | 37.0 | 7.0 | 3.62 | 106.0 | 167.0 | 316.0 | 37.0 | ... | 0.071588 | 0.388904 | 1.177423 | 1.290403 | 1.978475 | 0.039736 | 0.926851 | 0.092434 | 1.746121 | 0.459282 |
5 rows × 187 columns
In [4]:
import seaborn as sns
In [5]:
out_cols = ['PM2.5', 'PM10', 'SO2', 'NO2', 'O3', 'CO'] feature_cols = [x for x in data.columns if x not in out_cols and x != 'date'] feature_cols = feature_cols[144:] len(feature_cols), len(out_cols)
Out[5]:
(37, 6)
In [6]:
import tensorflow as tf from tensorflow import keras from tensorflow.keras import layers import tensorflow.keras.backend as K
2023-03-30 09:43:47.876739: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0
In [7]:
class TransformerBlock(layers.Layer): def __init__(self, embed_dim, num_heads, ff_dim, name, rate=0.1): super().__init__() self.att = layers.MultiHeadAttention(num_heads=num_heads, key_dim=embed_dim, name=name) self.ffn = keras.Sequential( [layers.Dense(ff_dim, activation="relu"), layers.Dense(embed_dim),] ) self.layernorm1 = layers.LayerNormalization(epsilon=1e-6) self.layernorm2 = layers.LayerNormalization(epsilon=1e-6) self.dropout1 = layers.Dropout(rate) self.dropout2 = layers.Dropout(rate) def call(self, inputs, training): attn_output = self.att(inputs, inputs) attn_output = self.dropout1(attn_output, training=training) out1 = self.layernorm1(inputs + attn_output) ffn_output = self.ffn(out1) ffn_output = self.dropout2(ffn_output, training=training) return self.layernorm2(out1 + ffn_output)
In [8]:
from tensorflow.keras import Model
In [9]:
from tensorflow.keras.initializers import Constant
In [10]:
# Custom loss layer class CustomMultiLossLayer(layers.Layer): def __init__(self, nb_outputs=2, **kwargs): self.nb_outputs = nb_outputs self.is_placeholder = True super(CustomMultiLossLayer, self).__init__(**kwargs) def build(self, input_shape=None): # initialise log_vars self.log_vars = [] for i in range(self.nb_outputs): self.log_vars += [self.add_weight(name='log_var' + str(i), shape=(1,), initializer=tf.initializers.he_normal(), trainable=True)] super(CustomMultiLossLayer, self).build(input_shape) def multi_loss(self, ys_true, ys_pred): assert len(ys_true) == self.nb_outputs and len(ys_pred) == self.nb_outputs loss = 0 for y_true, y_pred, log_var in zip(ys_true, ys_pred, self.log_vars): mse = (y_true - y_pred) ** 2. pre = K.exp(-log_var[0]) loss += tf.abs(tf.reduce_logsumexp(pre * mse + log_var[0], axis=-1)) return K.mean(loss) def call(self, inputs): ys_true = inputs[:self.nb_outputs] ys_pred = inputs[self.nb_outputs:] loss = self.multi_loss(ys_true, ys_pred) self.add_loss(loss, inputs=inputs) # We won't actually use the output. return K.concatenate(inputs, -1)
In [11]:
num_heads, ff_dim = 3, 16
In [12]:
def get_prediction_model(): def build_output(out, out_name): self_block = TransformerBlock(64, num_heads, ff_dim, name=f'{out_name}_attn') out = self_block(out) out = layers.GlobalAveragePooling1D()(out) out = layers.Dropout(0.1)(out) out = layers.Dense(32, activation="relu")(out) # out = layers.Dense(1, name=out_name, activation="sigmoid")(out) return out inputs = layers.Input(shape=(1,len(feature_cols)), name='input') x = layers.Conv1D(filters=64, kernel_size=1, activation='relu')(inputs) # x = layers.Dropout(rate=0.1)(x) lstm_out = layers.Bidirectional(layers.LSTM(units=64, return_sequences=True))(x) lstm_out = layers.Dense(128, activation='relu')(lstm_out) transformer_block = TransformerBlock(128, num_heads, ff_dim, name='first_attn') out = transformer_block(lstm_out) out = layers.GlobalAveragePooling1D()(out) out = layers.Dropout(0.1)(out) out = layers.Dense(64, activation='relu')(out) out = K.expand_dims(out, axis=1) pm25 = build_output(out, 'pm25') pm10 = build_output(out, 'pm10') so2 = build_output(out, 'so2') no2 = build_output(out, 'no2') o3 = build_output(out, 'o3') co = build_output(out, 'co') merge = layers.Concatenate(axis=1)([pm25, pm10, so2, no2, o3, co]) merge = K.expand_dims(merge, axis=1) merge_attn = TransformerBlock(32*6, 3, 16, name='last_attn') out = merge_attn(merge) out = layers.GlobalAveragePooling1D()(out) out = layers.Dropout(0.1)(out) pm25 = layers.Dense(32, activation='relu')(out) pm10 = layers.Dense(32, activation='relu')(out) so2 = layers.Dense(32, activation='relu')(out) no2 = layers.Dense(32, activation='relu')(out) o3 = layers.Dense(32, activation='relu')(out) co = layers.Dense(32, activation='relu')(out) pm25 = layers.Dense(1, activation='sigmoid', name='pm25')(pm25) pm10 = layers.Dense(1, activation='sigmoid', name='pm10')(pm10) so2 = layers.Dense(1, activation='sigmoid', name='so2')(so2) no2 = layers.Dense(1, activation='sigmoid', name='no2')(no2) o3 = layers.Dense(1, activation='sigmoid', name='o3')(o3) co = layers.Dense(1, activation='sigmoid', name='co')(co) model = Model(inputs=[inputs], outputs=[pm25, pm10, so2, no2, o3, co]) return model def get_trainable_model(prediction_model): inputs = layers.Input(shape=(1,len(feature_cols)), name='input') pm25, pm10, so2, no2, o3, co = prediction_model(inputs) pm25_real = layers.Input(shape=(1,), name='pm25_real') pm10_real = layers.Input(shape=(1,), name='pm10_real') so2_real = layers.Input(shape=(1,), name='so2_real') no2_real = layers.Input(shape=(1,), name='no2_real') o3_real = layers.Input(shape=(1,), name='o3_real') co_real = layers.Input(shape=(1,), name='co_real') out = CustomMultiLossLayer(nb_outputs=6)([pm25_real, pm10_real, so2_real, no2_real, o3_real, co_real, pm25, pm10, so2, no2, o3, co]) return Model([inputs, pm25_real, pm10_real, so2_real, no2_real, o3_real, co_real], out)
In [13]:
use_cols = feature_cols + out_cols use_data = data[use_cols].dropna()
In [14]:
for col in use_cols: use_data[col] = use_data[col].astype('float32')
In [15]:
maxs = use_data.max() mins = use_data.min() use_cols = use_data.columns
In [16]:
for col in use_cols: # use_data[col] = use_data[col].apply(lambda x: 0 if x < 0 else x) # use_data[col] = np.log1p(use_data[col]) use_data[col] = (use_data[col] - mins[col]) / (maxs[col] - mins[col])
In [17]:
train_data, valid = train_test_split(use_data[use_cols], test_size=0.1, random_state=42, shuffle=True) valid_data, test_data = train_test_split(valid, test_size=0.5, random_state=42, shuffle=True)
In [18]:
prediction_model = get_prediction_model() trainable_model = get_trainable_model(prediction_model)
2023-03-30 09:43:54.452903: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcuda.so.1 2023-03-30 09:43:54.512341: E tensorflow/stream_executor/cuda/cuda_driver.cc:328] failed call to cuInit: CUDA_ERROR_INVALID_DEVICE: invalid device ordinal 2023-03-30 09:43:54.512378: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:169] retrieving CUDA diagnostic information for host: ubuntu-NF5468M6 2023-03-30 09:43:54.512384: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:176] hostname: ubuntu-NF5468M6 2023-03-30 09:43:54.512489: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:200] libcuda reported version is: 510.47.3 2023-03-30 09:43:54.512510: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:204] kernel reported version is: 510.47.3 2023-03-30 09:43:54.512515: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:310] kernel version seems to match DSO: 510.47.3 2023-03-30 09:43:54.512816: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 AVX512F FMA To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
In [19]:
prediction_model.summary()
Model: "model"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input (InputLayer) [(None, 1, 37)] 0
__________________________________________________________________________________________________
conv1d (Conv1D) (None, 1, 64) 2432 input[0][0]
__________________________________________________________________________________________________
bidirectional (Bidirectional) (None, 1, 128) 66048 conv1d[0][0]
__________________________________________________________________________________________________
dense (Dense) (None, 1, 128) 16512 bidirectional[0][0]
__________________________________________________________________________________________________
transformer_block (TransformerB (None, 1, 128) 202640 dense[0][0]
__________________________________________________________________________________________________
global_average_pooling1d (Globa (None, 128) 0 transformer_block[0][0]
__________________________________________________________________________________________________
dropout_2 (Dropout) (None, 128) 0 global_average_pooling1d[0][0]
__________________________________________________________________________________________________
dense_3 (Dense) (None, 64) 8256 dropout_2[0][0]
__________________________________________________________________________________________________
tf.expand_dims (TFOpLambda) (None, 1, 64) 0 dense_3[0][0]
__________________________________________________________________________________________________
transformer_block_1 (Transforme (None, 1, 64) 52176 tf.expand_dims[0][0]
__________________________________________________________________________________________________
transformer_block_2 (Transforme (None, 1, 64) 52176 tf.expand_dims[0][0]
__________________________________________________________________________________________________
transformer_block_3 (Transforme (None, 1, 64) 52176 tf.expand_dims[0][0]
__________________________________________________________________________________________________
transformer_block_4 (Transforme (None, 1, 64) 52176 tf.expand_dims[0][0]
__________________________________________________________________________________________________
transformer_block_5 (Transforme (None, 1, 64) 52176 tf.expand_dims[0][0]
__________________________________________________________________________________________________
transformer_block_6 (Transforme (None, 1, 64) 52176 tf.expand_dims[0][0]
__________________________________________________________________________________________________
global_average_pooling1d_1 (Glo (None, 64) 0 transformer_block_1[0][0]
__________________________________________________________________________________________________
global_average_pooling1d_2 (Glo (None, 64) 0 transformer_block_2[0][0]
__________________________________________________________________________________________________
global_average_pooling1d_3 (Glo (None, 64) 0 transformer_block_3[0][0]
__________________________________________________________________________________________________
global_average_pooling1d_4 (Glo (None, 64) 0 transformer_block_4[0][0]
__________________________________________________________________________________________________
global_average_pooling1d_5 (Glo (None, 64) 0 transformer_block_5[0][0]
__________________________________________________________________________________________________
global_average_pooling1d_6 (Glo (None, 64) 0 transformer_block_6[0][0]
__________________________________________________________________________________________________
dropout_5 (Dropout) (None, 64) 0 global_average_pooling1d_1[0][0]
__________________________________________________________________________________________________
dropout_8 (Dropout) (None, 64) 0 global_average_pooling1d_2[0][0]
__________________________________________________________________________________________________
dropout_11 (Dropout) (None, 64) 0 global_average_pooling1d_3[0][0]
__________________________________________________________________________________________________
dropout_14 (Dropout) (None, 64) 0 global_average_pooling1d_4[0][0]
__________________________________________________________________________________________________
dropout_17 (Dropout) (None, 64) 0 global_average_pooling1d_5[0][0]
__________________________________________________________________________________________________
dropout_20 (Dropout) (None, 64) 0 global_average_pooling1d_6[0][0]
__________________________________________________________________________________________________
dense_6 (Dense) (None, 32) 2080 dropout_5[0][0]
__________________________________________________________________________________________________
dense_9 (Dense) (None, 32) 2080 dropout_8[0][0]
__________________________________________________________________________________________________
dense_12 (Dense) (None, 32) 2080 dropout_11[0][0]
__________________________________________________________________________________________________
dense_15 (Dense) (None, 32) 2080 dropout_14[0][0]
__________________________________________________________________________________________________
dense_18 (Dense) (None, 32) 2080 dropout_17[0][0]
__________________________________________________________________________________________________
dense_21 (Dense) (None, 32) 2080 dropout_20[0][0]
__________________________________________________________________________________________________
concatenate (Concatenate) (None, 192) 0 dense_6[0][0]
dense_9[0][0]
dense_12[0][0]
dense_15[0][0]
dense_18[0][0]
dense_21[0][0]
__________________________________________________________________________________________________
tf.expand_dims_1 (TFOpLambda) (None, 1, 192) 0 concatenate[0][0]
__________________________________________________________________________________________________
transformer_block_7 (Transforme (None, 1, 192) 451408 tf.expand_dims_1[0][0]
__________________________________________________________________________________________________
global_average_pooling1d_7 (Glo (None, 192) 0 transformer_block_7[0][0]
__________________________________________________________________________________________________
dropout_23 (Dropout) (None, 192) 0 global_average_pooling1d_7[0][0]
__________________________________________________________________________________________________
dense_24 (Dense) (None, 32) 6176 dropout_23[0][0]
__________________________________________________________________________________________________
dense_25 (Dense) (None, 32) 6176 dropout_23[0][0]
__________________________________________________________________________________________________
dense_26 (Dense) (None, 32) 6176 dropout_23[0][0]
__________________________________________________________________________________________________
dense_27 (Dense) (None, 32) 6176 dropout_23[0][0]
__________________________________________________________________________________________________
dense_28 (Dense) (None, 32) 6176 dropout_23[0][0]
__________________________________________________________________________________________________
dense_29 (Dense) (None, 32) 6176 dropout_23[0][0]
__________________________________________________________________________________________________
pm25 (Dense) (None, 1) 33 dense_24[0][0]
__________________________________________________________________________________________________
pm10 (Dense) (None, 1) 33 dense_25[0][0]
__________________________________________________________________________________________________
so2 (Dense) (None, 1) 33 dense_26[0][0]
__________________________________________________________________________________________________
no2 (Dense) (None, 1) 33 dense_27[0][0]
__________________________________________________________________________________________________
o3 (Dense) (None, 1) 33 dense_28[0][0]
__________________________________________________________________________________________________
co (Dense) (None, 1) 33 dense_29[0][0]
==================================================================================================
Total params: 1,110,086
Trainable params: 1,110,086
Non-trainable params: 0
__________________________________________________________________________________________________
In [20]:
from tensorflow.keras import optimizers
In [21]:
trainable_model.compile(optimizer='adam', loss=None)
In [22]:
X = np.expand_dims(train_data[feature_cols].values, axis=1) Y = [x for x in train_data[out_cols].values.T] Y_valid = [x for x in valid_data[out_cols].values.T]
In [23]:
from keras.callbacks import ReduceLROnPlateau reduce_lr = ReduceLROnPlateau(monitor='val_loss', patience=10, mode='auto')
In [24]:
trainable_model.compile(optimizer='adam', loss=None) hist = trainable_model.fit([X, Y[0], Y[1], Y[2], Y[3], Y[4], Y[5]], epochs=100, batch_size=64, verbose=1, validation_data=[np.expand_dims(valid_data[feature_cols].values, axis=1), Y_valid[0], Y_valid[1], Y_valid[2], Y_valid[3], Y_valid[4], Y_valid[5]], callbacks=[reduce_lr] )
2023-03-30 09:44:00.730399: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:176] None of the MLIR Optimization Passes are enabled (registered 2) 2023-03-30 09:44:00.750292: I tensorflow/core/platform/profile_utils/cpu_utils.cc:114] CPU Frequency: 2200000000 Hz
Epoch 1/100 690/690 [==============================] - 22s 20ms/step - loss: 2.0729 - val_loss: 1.2554 Epoch 2/100 690/690 [==============================] - 12s 18ms/step - loss: 0.7311 - val_loss: 0.3496 Epoch 3/100 690/690 [==============================] - 12s 18ms/step - loss: 0.1033 - val_loss: 0.0213 Epoch 4/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0228 - val_loss: 0.0188 Epoch 5/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0224 - val_loss: 0.0237 Epoch 6/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0217 - val_loss: 0.0196 Epoch 7/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0210 - val_loss: 0.0249 Epoch 8/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0214 - val_loss: 0.0212 Epoch 9/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0202 - val_loss: 0.0181 Epoch 10/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0197 - val_loss: 0.0221 Epoch 11/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0197 - val_loss: 0.0189 Epoch 12/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0191 - val_loss: 0.0176 Epoch 13/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0188 - val_loss: 0.0180 Epoch 14/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0185 - val_loss: 0.0176 Epoch 15/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0183 - val_loss: 0.0177 Epoch 16/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0181 - val_loss: 0.0167 Epoch 17/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0182 - val_loss: 0.0171 Epoch 18/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0177 - val_loss: 0.0169 Epoch 19/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0180 - val_loss: 0.0164 Epoch 20/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0173 - val_loss: 0.0163 Epoch 21/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0168 - val_loss: 0.0167 Epoch 22/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0172 - val_loss: 0.0167 Epoch 23/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0169 - val_loss: 0.0151 Epoch 24/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0169 - val_loss: 0.0178 Epoch 25/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0166 - val_loss: 0.0156 Epoch 26/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0163 - val_loss: 0.0170 Epoch 27/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0162 - val_loss: 0.0151 Epoch 28/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0167 - val_loss: 0.0147 Epoch 29/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0165 - val_loss: 0.0155 Epoch 30/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0159 - val_loss: 0.0152 Epoch 31/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0156 - val_loss: 0.0149 Epoch 32/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0161 - val_loss: 0.0150 Epoch 33/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0153 - val_loss: 0.0148 Epoch 34/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0156 - val_loss: 0.0160 Epoch 35/100 690/690 [==============================] - 13s 18ms/step - loss: 0.0153 - val_loss: 0.0164 Epoch 36/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0151 - val_loss: 0.0147 Epoch 37/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0152 - val_loss: 0.0146 Epoch 38/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0152 - val_loss: 0.0144 Epoch 39/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0147 - val_loss: 0.0142 Epoch 40/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0149 - val_loss: 0.0141 Epoch 41/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0147 - val_loss: 0.0156 Epoch 42/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0145 - val_loss: 0.0144 Epoch 43/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0146 - val_loss: 0.0135 Epoch 44/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0143 - val_loss: 0.0184 Epoch 45/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0147 - val_loss: 0.0142 Epoch 46/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0144 - val_loss: 0.0136 Epoch 47/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0143 - val_loss: 0.0134 Epoch 48/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0139 - val_loss: 0.0136 Epoch 49/100 690/690 [==============================] - 12s 17ms/step - loss: 0.0143 - val_loss: 0.0151 Epoch 50/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0139 - val_loss: 0.0144 Epoch 51/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0141 - val_loss: 0.0144 Epoch 52/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0135 - val_loss: 0.0136 Epoch 53/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0136 - val_loss: 0.0137 Epoch 54/100 690/690 [==============================] - 13s 18ms/step - loss: 0.0118 - val_loss: 0.0119 Epoch 55/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0115 - val_loss: 0.0118 Epoch 56/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0114 - val_loss: 0.0118 Epoch 57/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0114 - val_loss: 0.0117 Epoch 58/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0113 - val_loss: 0.0118 Epoch 59/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0112 - val_loss: 0.0118 Epoch 60/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0112 - val_loss: 0.0117 Epoch 61/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0112 - val_loss: 0.0118 Epoch 62/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0111 - val_loss: 0.0117 Epoch 63/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0111 - val_loss: 0.0117 Epoch 64/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0111 - val_loss: 0.0116 Epoch 65/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0110 - val_loss: 0.0115 Epoch 66/100 690/690 [==============================] - 13s 18ms/step - loss: 0.0110 - val_loss: 0.0117 Epoch 67/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0110 - val_loss: 0.0116 Epoch 68/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0110 - val_loss: 0.0116 Epoch 69/100 690/690 [==============================] - 13s 18ms/step - loss: 0.0109 - val_loss: 0.0116 Epoch 70/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0110 - val_loss: 0.0114 Epoch 71/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0109 - val_loss: 0.0115 Epoch 72/100 690/690 [==============================] - 13s 18ms/step - loss: 0.0109 - val_loss: 0.0116 Epoch 73/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0108 - val_loss: 0.0115 Epoch 74/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0108 - val_loss: 0.0115 Epoch 75/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0108 - val_loss: 0.0116 Epoch 76/100 690/690 [==============================] - 13s 18ms/step - loss: 0.0108 - val_loss: 0.0114 Epoch 77/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0107 - val_loss: 0.0115 Epoch 78/100 690/690 [==============================] - 13s 18ms/step - loss: 0.0107 - val_loss: 0.0115 Epoch 79/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0107 - val_loss: 0.0115 Epoch 80/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0107 - val_loss: 0.0115 Epoch 81/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0105 - val_loss: 0.0113 Epoch 82/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0104 - val_loss: 0.0113 Epoch 83/100 690/690 [==============================] - 14s 20ms/step - loss: 0.0104 - val_loss: 0.0114 Epoch 84/100 690/690 [==============================] - 14s 21ms/step - loss: 0.0104 - val_loss: 0.0113 Epoch 85/100 690/690 [==============================] - 13s 19ms/step - loss: 0.0104 - val_loss: 0.0113 Epoch 86/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0104 - val_loss: 0.0113 Epoch 87/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0104 - val_loss: 0.0113 Epoch 88/100 690/690 [==============================] - 13s 18ms/step - loss: 0.0104 - val_loss: 0.0113 Epoch 89/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0104 - val_loss: 0.0113 Epoch 90/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0104 - val_loss: 0.0113 Epoch 91/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0104 - val_loss: 0.0113 Epoch 92/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0103 - val_loss: 0.0113 Epoch 93/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0103 - val_loss: 0.0113 Epoch 94/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0104 - val_loss: 0.0113 Epoch 95/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0103 - val_loss: 0.0113 Epoch 96/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0103 - val_loss: 0.0113 Epoch 97/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0103 - val_loss: 0.0113 Epoch 98/100 690/690 [==============================] - 12s 17ms/step - loss: 0.0103 - val_loss: 0.0113 Epoch 99/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0103 - val_loss: 0.0113 Epoch 100/100 690/690 [==============================] - 12s 18ms/step - loss: 0.0104 - val_loss: 0.0113
In [25]:
prediction_model.save(f'./models/uw_loss_looknow/')
WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.
2023-03-30 10:05:02.545274: W tensorflow/python/util/util.cc:348] Sets are not currently considered sequences, but this may change in the future, so consider avoiding using them. WARNING:absl:Found untraced functions such as first_attn_layer_call_fn, first_attn_layer_call_and_return_conditional_losses, layer_normalization_layer_call_fn, layer_normalization_layer_call_and_return_conditional_losses, layer_normalization_1_layer_call_fn while saving (showing 5 of 450). These functions will not be directly callable after loading.
INFO:tensorflow:Assets written to: ./models/uw_loss_looknow/assets
INFO:tensorflow:Assets written to: ./models/uw_loss_looknow/assets
In [26]:
rst = prediction_model.predict(np.expand_dims(test_data[feature_cols], axis=1)) rst
Out[26]:
[array([[0.05530462],
[0.15461421],
[0.06984487],
...,
[0.11500913],
[0.0667375 ],
[0.2275947 ]], dtype=float32),
array([[0.04914668],
[0.10565668],
[0.0540773 ],
...,
[0.12981808],
[0.06497446],
[0.13074341]], dtype=float32),
array([[0.0275037 ],
[0.03192595],
[0.02210939],
...,
[0.04022893],
[0.02630579],
[0.04132178]], dtype=float32),
array([[0.06796426],
[0.20642754],
[0.28993258],
...,
[0.4715118 ],
[0.09288657],
[0.5116445 ]], dtype=float32),
array([[0.25186226],
[0.35088968],
[0.09913486],
...,
[0.03776854],
[0.47971994],
[0.05172443]], dtype=float32),
array([[0.06528944],
[0.12970403],
[0.08967546],
...,
[0.10639769],
[0.0641529 ],
[0.18867406]], dtype=float32)]
In [27]:
[np.exp(K.get_value(log_var[0]))**0.5 for log_var in trainable_model.layers[-1].log_vars]
Out[27]:
[0.9995770752172775, 0.9998096046393907, 0.9998504109554951, 0.9992896477643191, 0.9997416699046467, 0.9996395653611235]
In [28]:
pred_rst = pd.DataFrame.from_records(np.squeeze(np.asarray(rst), axis=2).T, columns=out_cols)
In [29]:
real_rst = test_data[out_cols].copy()
In [30]:
for col in out_cols: pred_rst[col] = pred_rst[col] * (maxs[col] - mins[col]) + mins[col] real_rst[col] = real_rst[col] * (maxs[col] - mins[col]) + mins[col]
In [31]:
y_pred_pm25 = pred_rst['PM2.5'].values.reshape(-1,) y_pred_pm10 = pred_rst['PM10'].values.reshape(-1,) y_pred_so2 = pred_rst['SO2'].values.reshape(-1,) y_pred_no2 = pred_rst['NO2'].values.reshape(-1,) y_pred_o3 = pred_rst['O3'].values.reshape(-1,) y_pred_co = pred_rst['CO'].values.reshape(-1,) y_true_pm25 = real_rst['PM2.5'].values.reshape(-1,) y_true_pm10 = real_rst['PM10'].values.reshape(-1,) y_true_so2 = real_rst['SO2'].values.reshape(-1,) y_true_no2 = real_rst['NO2'].values.reshape(-1,) y_true_o3 = real_rst['O3'].values.reshape(-1,) y_true_co = real_rst['CO'].values.reshape(-1,)
In [32]:
from sklearn.metrics import mean_squared_error, mean_absolute_error, r2_score, mean_absolute_percentage_error
In [33]:
def print_eva(y_true, y_pred, tp): MSE = mean_squared_error(y_true, y_pred) RMSE = np.sqrt(MSE) MAE = mean_absolute_error(y_true, y_pred) MAPE = mean_absolute_percentage_error(y_true, y_pred) R_2 = r2_score(y_true, y_pred) print(f"COL: {tp}, MSE: {format(MSE, '.2E')}", end=',') print(f'RMSE: {round(RMSE, 4)}', end=',') print(f'MAPE: {round(MAPE, 4) * 100} %', end=',') print(f'MAE: {round(MAE, 4)}', end=',') print(f'R_2: {round(R_2, 4)}') return [MSE, RMSE, MAE, MAPE, R_2]
In [34]:
pm25_eva = print_eva(y_true_pm25, y_pred_pm25, tp='pm25') pm10_eva = print_eva(y_true_pm10, y_pred_pm10, tp='pm10') so2_eva = print_eva(y_true_so2, y_pred_so2, tp='so2') nox_eva = print_eva(y_true_no2, y_pred_no2, tp='no2') o3_eva = print_eva(y_true_o3, y_pred_o3, tp='o3') co_eva = print_eva(y_true_co, y_pred_co, tp='co')
COL: pm25, MSE: 6.07E+02,RMSE: 24.63920021057129,MAPE: 47.06000089645386 %,MAE: 17.84910011291504,R_2: 0.7534 COL: pm10, MSE: 2.18E+03,RMSE: 46.695701599121094,MAPE: 38.22999894618988 %,MAE: 32.543701171875,R_2: 0.6114 COL: so2, MSE: 6.40E+02,RMSE: 25.30109977722168,MAPE: 69.72000002861023 %,MAE: 15.265700340270996,R_2: 0.7953 COL: no2, MSE: 1.25E+02,RMSE: 11.198399543762207,MAPE: 23.669999837875366 %,MAE: 8.1274995803833,R_2: 0.7998 COL: o3, MSE: 1.39E+02,RMSE: 11.781700134277344,MAPE: 45.170000195503235 %,MAE: 8.724499702453613,R_2: 0.9443 COL: co, MSE: 9.53E-02,RMSE: 0.30869999527931213,MAPE: 21.150000393390656 %,MAE: 0.22669999301433563,R_2: 0.8096
In [35]:
pd.DataFrame.from_records([pm25_eva, pm10_eva, so2_eva, nox_eva, o3_eva,co_eva], columns=['MSE', 'RMSE', 'MAE', 'MAPE', 'R_2'], index=['PM25', 'PM10', 'SO2', 'NO2', 'O3', 'CO'])
Out[35]:
| MSE | RMSE | MAE | MAPE | R_2 | |
|---|---|---|---|---|---|
| PM25 | 607.088562 | 24.639168 | 17.849148 | 0.470577 | 0.753417 |
| PM10 | 2180.489258 | 46.695709 | 32.543747 | 0.382252 | 0.611418 |
| SO2 | 640.143799 | 25.301064 | 15.265686 | 0.697226 | 0.795343 |
| NO2 | 125.403076 | 11.198352 | 8.127537 | 0.236683 | 0.799830 |
| O3 | 138.808472 | 11.781701 | 8.724504 | 0.451706 | 0.944349 |
| CO | 0.095273 | 0.308662 | 0.226732 | 0.211459 | 0.809550 |
In [36]:
plt.figure(figsize=(16, 9)) plt.plot(pred_rst['PM10'].values[50:150], 'o-', label='pred') plt.plot(real_rst['PM10'].values[50:150], '*-', label='real') plt.legend(loc='best')
Out[36]:
<matplotlib.legend.Legend at 0x7f9e107bdc90>
WARNING:matplotlib.font_manager:findfont: Font family ['sans-serif'] not found. Falling back to DejaVu Sans. WARNING:matplotlib.font_manager:findfont: Generic family 'sans-serif' not found because none of the following families were found: SimHei
In [38]:
from statistics import mean import matplotlib.pyplot as plt from sklearn.metrics import explained_variance_score,r2_score, median_absolute_error, mean_squared_error, mean_absolute_error from scipy import stats import numpy as np from matplotlib import rcParams config = {"font.size": 32,"mathtext.fontset":'stix'} rcParams.update(config)
In [39]:
config = {"font.size": 32,"mathtext.fontset":'stix'} rcParams.update(config)
In [40]:
def scatter_out_1(x, y, label, name): ## x,y为两个需要做对比分析的两个量。 # ==========计算评价指标========== BIAS = mean(x - y) MSE = mean_squared_error(x, y) RMSE = np.power(MSE, 0.5) R2 = r2_score(x, y) MAE = mean_absolute_error(x, y) EV = explained_variance_score(x, y) print('==========算法评价指标==========') print('Explained Variance(EV):', '%.3f' % (EV)) print('Mean Absolute Error(MAE):', '%.3f' % (MAE)) print('Mean squared error(MSE):', '%.3f' % (MSE)) print('Root Mean Squard Error(RMSE):', '%.3f' % (RMSE)) print('R_squared:', '%.3f' % (R2)) # ===========Calculate the point density========== xy = np.vstack([x, y]) z = stats.gaussian_kde(xy)(xy) # ===========Sort the points by density, so that the densest points are plotted last=========== idx = z.argsort() x, y, z = x[idx], y[idx], z[idx] def best_fit_slope_and_intercept(xs, ys): m = (((mean(xs) * mean(ys)) - mean(xs * ys)) / ((mean(xs) * mean(xs)) - mean(xs * xs))) b = mean(ys) - m * mean(xs) return m, b m, b = best_fit_slope_and_intercept(x, y) regression_line = [] for a in x: regression_line.append((m * a) + b) fig,ax=plt.subplots(figsize=(12,9),dpi=400) scatter=ax.scatter(x,y,marker='o',c=z*100,s=15,label='LST',cmap='Spectral_r') cbar=plt.colorbar(scatter,shrink=1,orientation='vertical',extend='both',pad=0.015,aspect=30,label='Frequency') min_value = min(min(x), min(y)) max_value = max(max(x), max(y)) plt.plot([min_value-5,max_value+5],[min_value-5,max_value+5],'black',lw=1.5) # 画的1:1线，线的颜色为black，线宽为0.8 plt.plot(x,regression_line,'red',lw=1.5) # 预测与实测数据之间的回归线 plt.axis([min_value-5,max_value+5,min_value-5,max_value+5]) # 设置线的范围 plt.xlabel('Measured %s' % label) plt.ylabel('Retrived %s' % label) # plt.xticks(fontproperties='Times New Roman') # plt.yticks(fontproperties='Times New Roman') plt.text(min_value-5 + (max_value-min_value) * 0.05, int(max_value * 0.95), '$N=%.f$' % len(y)) # text的位置需要根据x,y的大小范围进行调整。 plt.text(min_value-5 + (max_value-min_value) * 0.05, int(max_value * 0.88), '$R^2=%.2f$' % R2) plt.text(min_value-5 + (max_value-min_value) * 0.05, int(max_value * 0.81), '$RMSE=%.2f$' % RMSE) plt.xlim(min_value-5,max_value+5) # 设置x坐标轴的显示范围 plt.ylim(min_value-5,max_value+5) # 设置y坐标轴的显示范围 # file_name = name.split('(')[0].strip() plt.savefig(f'./figure/looknow/{name}.png',dpi=800,bbox_inches='tight',pad_inches=0) plt.show()
In [41]:
scatter_out_1(real_rst['PM10'].values, pred_rst['PM10'].values, label='$PM_{10}\ (\mu g/m^3$)', name='PM10')
==========算法评价指标========== Explained Variance(EV): 0.611 Mean Absolute Error(MAE): 32.544 Mean squared error(MSE): 2180.489 Root Mean Squard Error(RMSE): 46.696 R_squared: 0.611
WARNING:matplotlib.font_manager:findfont: Font family ['sans-serif'] not found. Falling back to DejaVu Sans. WARNING:matplotlib.font_manager:findfont: Generic family 'sans-serif' not found because none of the following families were found: SimHei
In [42]:
scatter_out_1(real_rst['PM2.5'].values, pred_rst['PM2.5'].values, label='$PM_{2.5} (\mu g/m^3$)', name='PM25')
==========算法评价指标========== Explained Variance(EV): 0.754 Mean Absolute Error(MAE): 17.849 Mean squared error(MSE): 607.089 Root Mean Squard Error(RMSE): 24.639 R_squared: 0.753
In [43]:
scatter_out_1(real_rst['SO2'].values, pred_rst['SO2'].values, label='$SO_2\ (\mu g/m^3)$', name='SO2')
==========算法评价指标========== Explained Variance(EV): 0.797 Mean Absolute Error(MAE): 15.266 Mean squared error(MSE): 640.144 Root Mean Squard Error(RMSE): 25.301 R_squared: 0.795
In [44]:
scatter_out_1(real_rst['NO2'].values, pred_rst['NO2'].values, label='$NO_2\ (\mu g/m^3)$', name='NO2')
==========算法评价指标========== Explained Variance(EV): 0.800 Mean Absolute Error(MAE): 8.128 Mean squared error(MSE): 125.403 Root Mean Squard Error(RMSE): 11.198 R_squared: 0.800
In [45]:
scatter_out_1(real_rst['O3'], pred_rst['O3'], label='$O_3 \ (\mu g/m^3)$', name='O3')
==========算法评价指标========== Explained Variance(EV): 0.944 Mean Absolute Error(MAE): 8.725 Mean squared error(MSE): 138.808 Root Mean Squard Error(RMSE): 11.782 R_squared: 0.944
In [46]:
def scatter_out_2(x, y, name): ## x,y为两个需要做对比分析的两个量。 # ==========计算评价指标========== BIAS = mean(x - y) MSE = mean_squared_error(x, y) RMSE = np.power(MSE, 0.5) R2 = r2_score(x, y) MAE = mean_absolute_error(x, y) EV = explained_variance_score(x, y) print('==========算法评价指标==========') print('Explained Variance(EV):', '%.3f' % (EV)) print('Mean Absolute Error(MAE):', '%.3f' % (MAE)) print('Mean squared error(MSE):', '%.3f' % (MSE)) print('Root Mean Squard Error(RMSE):', '%.3f' % (RMSE)) print('R_squared:', '%.3f' % (R2)) # ===========Calculate the point density========== xy = np.vstack([x, y]) z = stats.gaussian_kde(xy)(xy) # ===========Sort the points by density, so that the densest points are plotted last=========== idx = z.argsort() x, y, z = x[idx], y[idx], z[idx] def best_fit_slope_and_intercept(xs, ys): m = (((mean(xs) * mean(ys)) - mean(xs * ys)) / ((mean(xs) * mean(xs)) - mean(xs * xs))) b = mean(ys) - m * mean(xs) return m, b m, b = best_fit_slope_and_intercept(x, y) regression_line = [] for a in x: regression_line.append((m * a) + b) fig,ax=plt.subplots(figsize=(12,9),dpi=400) scatter=ax.scatter(x,y,marker='o',c=z*100,s=15,label='LST',cmap='Spectral_r') cbar=plt.colorbar(scatter,shrink=1,orientation='vertical',extend='both',pad=0.015,aspect=30,label='frequency') plt.plot([0, 6], [0, 6],'black',lw=1.5) # 画的1:1线，线的颜色为black，线宽为0.8 plt.plot(x,regression_line,'red',lw=1.5) # 预测与实测数据之间的回归线 plt.axis([0,6,0,6]) # 设置线的范围 plt.xlabel(f'Measured {name}') plt.ylabel(f'Retrived {name}') # plt.xticks(fontproperties='Times New Roman') # plt.yticks(fontproperties='Times New Roman') plt.text(0.3, 5.5, '$N=%.f$' % len(y)) # text的位置需要根据x,y的大小范围进行调整。 plt.text(0.3, 5.0, '$R^2=%.2f$' % R2) plt.text(0.3, 4.5, '$RMSE=%.2f$' % RMSE) plt.xlim(0, 6) # 设置x坐标轴的显示范围 plt.ylim(0, 6) # 设置y坐标轴的显示范围 file_name = name.split('(')[0].strip() plt.savefig(f'./figure/looknow/CO.png',dpi=800,bbox_inches='tight',pad_inches=0) plt.show()
In [47]:
scatter_out_2(real_rst['CO'].values, pred_rst['CO'].values, name='$CO \ (mg/m^3)$')
==========算法评价指标========== Explained Variance(EV): 0.810 Mean Absolute Error(MAE): 0.227 Mean squared error(MSE): 0.095 Root Mean Squard Error(RMSE): 0.309 R_squared: 0.810
In [ ]: