R2 LSTM-火灾温度预测

月亮泡面=

于 2024-03-20 11:35:32 发布

阅读量610

点赞数 16

文章标签： lstm

本文链接：https://blog.csdn.net/weixin_45103399/article/details/136871573

版权

🍨 本文为🔗365天深度学习训练营中的学习记录博客
🍖 原作者：K同学啊 | 接辅导、项目定制
🚀 文章来源：K同学的学习圈子

我的环境：

语言环境：python 3.8
编译器：jupyter notebook
深度学习环境：Pytorch

torch == 2.1.0+cpu

torchvision == 0.16.0+cpu

LSTM介绍

在这里插入图片描述

1. 输入输出

输入：当前时刻网络输出x_t，上一时刻LSTM输出h_t-1，上一时刻记忆单元c_t-1
输出：当前时刻LSTM输出h_t，隐藏状态向量h_t，记忆单元c_t

2. 三个门

遗忘门ft，表示h_t-1的那些特征被用于计算c_t，sigmoid函数
输入门it，决定哪些新信息被添加到记忆单元，sigmoid函数筛选信息*tanh生成候选信息
输出门ot，决定记忆单元哪些信息输出到h_t中，sigmoid筛选细腻*tanh处理后的记忆单元 = h_t

任务说明：

数据集提供了火灾温度（tem1），一氧化碳浓度（co1），烟雾浓度（soot1）随时间变化的数据，对未来某一时刻火灾温度进行预测。
R2达到0.83.
使用1-8时刻预测9-10时刻。

一、准备工作

import tensorflow as tf

import pandas as pd
import numpy as np

df = pd.read_csv("woodpine2.csv")

df

	Time	Tem1	CO 1	Soot 1
0	0.000	25.0	0.000000	0.000000
1	0.228	25.0	0.000000	0.000000
2	0.456	25.0	0.000000	0.000000
3	0.685	25.0	0.000000	0.000000
4	0.913	25.0	0.000000	0.000000
...	...	...	...	...
5943	366.000	295.0	0.000077	0.000496
5944	366.000	294.0	0.000077	0.000494
5945	367.000	292.0	0.000077	0.000491
5946	367.000	291.0	0.000076	0.000489
5947	367.000	290.0	0.000076	0.000487

5948 rows × 4 columns

plt.rcParams['savefig.dpi'] = 500 #像素
plt.rcParams['figure.dpi'] = 500 #分辨率

fig, ax = plt.subplots(1,3,constrained_layout = True, figsize = (14,3))

sns.lineplot(data = df["Tem1"], ax = ax[0])
sns.lineplot(data = df["CO 1"], ax = ax[1])
sns.lineplot(data = df["Soot 1"], ax = ax[2])

<Axes: ylabel='Soot 1'>

在这里插入图片描述

二、数据预处理

dataFrame = df.iloc[:,1:]
dataFrame

	Tem1	CO 1	Soot 1
0	25.0	0.000000	0.000000
1	25.0	0.000000	0.000000
2	25.0	0.000000	0.000000
3	25.0	0.000000	0.000000
4	25.0	0.000000	0.000000
...	...	...	...
5943	295.0	0.000077	0.000496
5944	294.0	0.000077	0.000494
5945	292.0	0.000077	0.000491
5946	291.0	0.000076	0.000489
5947	290.0	0.000076	0.000487

5948 rows × 3 columns

1. 设置X，y

width_X = 8 #前8个时段为X
width_y = 2 #第9个时段为y

X = []
y = []
in_start = 0

for _,_ in df.iterrows():
    in_end = in_start + width_X
    out_end = in_end + width_y

    if out_end < len(dataFrame):
        X_ = np.array(dataFrame.iloc[in_start:in_end , ])
        X_ = X_.reshape((len(X_)*3))
        y_ = np.array(dataFrame.iloc[in_end: out_end , 0])
        X.append(X_)
        y.append(y_)
    in_start += 1

X = np.array(X)
y = np.array(y)

X,y

(array([[2.50e+01, 0.00e+00, 0.00e+00, ..., 2.50e+01, 2.07e-38, 1.33e-37],
        [2.50e+01, 0.00e+00, 0.00e+00, ..., 2.50e+01, 5.37e-37, 3.45e-36],
        [2.50e+01, 0.00e+00, 0.00e+00, ..., 2.50e+01, 5.51e-36, 3.54e-35],
        ...,
        [2.96e+02, 7.84e-05, 5.03e-04, ..., 2.96e+02, 7.77e-05, 4.99e-04],
        [2.97e+02, 7.86e-05, 5.04e-04, ..., 2.95e+02, 7.73e-05, 4.96e-04],
        [2.97e+02, 7.87e-05, 5.05e-04, ..., 2.94e+02, 7.69e-05, 4.94e-04]]),
 array([[ 25.,  25.],
        [ 25.,  25.],
        [ 25.,  25.],
        ...,
        [295., 294.],
        [294., 292.],
        [292., 291.]]))

X.shape, y.shape

((5938, 24), (5938, 2))

2. 归一化

from sklearn.preprocessing import MinMaxScaler

sc = MinMaxScaler(feature_range = (0,1))
X_scales = sc.fit_transform(X)
X_scales.shape

(5938, 24)

X_scaled = X_scales.reshape(len(X_scales),width_X,3)
X_scaled.shape

(5938, 8, 3)

3. 划分数据集

X_train = np.array(X_scaled[:5000]).astype('float64')
y_train = np.array(y[:5000]).astype('float64')
X_test = np.array(X_scaled[5000:]).astype('float64')
y_test = np.array(y[5000:]).astype('float64')

X_train.shape, X_test.shape

((5000, 8, 3), (938, 8, 3))

三、模型构建，编译，训练

1. 模型构建

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, LSTM, Bidirectional
from tensorflow.keras        import Input

model = Sequential()
model.add(LSTM(units=64, activation = 'relu',return_sequences = True, input_shape = (X_train.shape[1],3)))
model.add(LSTM(units=64, activation = 'relu'))
model.add(Dense(width_y))
model.summary()

Model: "sequential_2"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 lstm_4 (LSTM)               (None, 8, 64)             17408     
                                                                 
 lstm_5 (LSTM)               (None, 64)                33024     
                                                                 
 dense_2 (Dense)             (None, 2)                 130       
                                                                 
=================================================================
Total params: 50562 (197.51 KB)
Trainable params: 50562 (197.51 KB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________

2. 模型编译

model.compile(optimizer = tf.keras.optimizers.Adam(1e-3),
              loss = 'mean_squared_error')

3. 模型训练

epochs = 50

history = model.fit(X_train,y_train,
                    epochs = epochs,
                    batch_size = 64,
                    validation_data = (X_test,y_test),
                    validation_freq = 1)

Epoch 1/50
79/79 [==============================] - 3s 15ms/step - loss: 13002.8154 - val_loss: 5616.8252
Epoch 2/50
79/79 [==============================] - 1s 12ms/step - loss: 247.7659 - val_loss: 1137.9130
Epoch 3/50
79/79 [==============================] - 1s 12ms/step - loss: 38.8453 - val_loss: 379.3899
Epoch 4/50
79/79 [==============================] - 1s 12ms/step - loss: 11.2683 - val_loss: 248.6566
Epoch 5/50
79/79 [==============================] - 1s 13ms/step - loss: 8.5500 - val_loss: 252.0033
Epoch 6/50
79/79 [==============================] - 1s 13ms/step - loss: 8.6463 - val_loss: 186.1750
Epoch 7/50
79/79 [==============================] - 1s 12ms/step - loss: 8.5890 - val_loss: 203.9141
Epoch 8/50
79/79 [==============================] - 1s 13ms/step - loss: 8.2541 - val_loss: 226.8261
Epoch 9/50
79/79 [==============================] - 1s 13ms/step - loss: 8.2206 - val_loss: 214.3781
Epoch 10/50
79/79 [==============================] - 1s 13ms/step - loss: 8.3749 - val_loss: 164.7413
Epoch 11/50
79/79 [==============================] - 1s 12ms/step - loss: 8.4348 - val_loss: 167.7527
Epoch 12/50
79/79 [==============================] - 1s 12ms/step - loss: 8.6248 - val_loss: 165.3832
Epoch 13/50
79/79 [==============================] - 1s 13ms/step - loss: 8.5666 - val_loss: 224.6095
Epoch 14/50
79/79 [==============================] - 1s 13ms/step - loss: 8.3841 - val_loss: 213.7288
Epoch 15/50
79/79 [==============================] - 1s 15ms/step - loss: 8.2504 - val_loss: 271.5983
Epoch 16/50
79/79 [==============================] - 1s 14ms/step - loss: 8.7495 - val_loss: 180.2835
Epoch 17/50
79/79 [==============================] - 1s 12ms/step - loss: 8.2108 - val_loss: 239.1326
Epoch 18/50
79/79 [==============================] - 1s 14ms/step - loss: 8.6858 - val_loss: 152.2928
Epoch 19/50
79/79 [==============================] - 1s 14ms/step - loss: 8.4797 - val_loss: 255.9244
Epoch 20/50
79/79 [==============================] - 1s 13ms/step - loss: 8.3837 - val_loss: 190.1214
Epoch 21/50
79/79 [==============================] - 1s 13ms/step - loss: 8.2621 - val_loss: 131.4390
Epoch 22/50
79/79 [==============================] - 1s 12ms/step - loss: 8.1924 - val_loss: 194.1792
Epoch 23/50
79/79 [==============================] - 1s 12ms/step - loss: 8.0023 - val_loss: 214.8212
Epoch 24/50
79/79 [==============================] - 1s 13ms/step - loss: 8.0245 - val_loss: 150.9851
Epoch 25/50
79/79 [==============================] - 1s 12ms/step - loss: 8.4004 - val_loss: 123.8255
Epoch 26/50
79/79 [==============================] - 1s 12ms/step - loss: 8.6597 - val_loss: 191.1270
Epoch 27/50
79/79 [==============================] - 1s 13ms/step - loss: 8.4472 - val_loss: 151.8523
Epoch 28/50
79/79 [==============================] - 1s 13ms/step - loss: 8.1072 - val_loss: 126.9261
Epoch 29/50
79/79 [==============================] - 1s 13ms/step - loss: 9.8221 - val_loss: 249.2247
Epoch 30/50
79/79 [==============================] - 1s 13ms/step - loss: 8.1813 - val_loss: 158.8967
Epoch 31/50
79/79 [==============================] - 1s 13ms/step - loss: 8.8521 - val_loss: 121.6337
Epoch 32/50
79/79 [==============================] - 1s 14ms/step - loss: 8.6561 - val_loss: 108.4152
Epoch 33/50
79/79 [==============================] - 1s 12ms/step - loss: 9.1036 - val_loss: 121.5477
Epoch 34/50
79/79 [==============================] - 1s 13ms/step - loss: 8.8073 - val_loss: 97.2747
Epoch 35/50
79/79 [==============================] - 1s 13ms/step - loss: 9.1284 - val_loss: 195.3769
Epoch 36/50
79/79 [==============================] - 1s 13ms/step - loss: 7.7126 - val_loss: 141.8356
Epoch 37/50
79/79 [==============================] - 1s 13ms/step - loss: 7.7963 - val_loss: 211.4055
Epoch 38/50
79/79 [==============================] - 1s 13ms/step - loss: 8.8638 - val_loss: 204.0073
Epoch 39/50
79/79 [==============================] - 1s 12ms/step - loss: 7.9304 - val_loss: 116.1591
Epoch 40/50
79/79 [==============================] - 1s 14ms/step - loss: 7.6471 - val_loss: 235.9073
Epoch 41/50
79/79 [==============================] - 1s 12ms/step - loss: 7.7511 - val_loss: 120.2813
Epoch 42/50
79/79 [==============================] - 1s 12ms/step - loss: 9.3354 - val_loss: 280.3375
Epoch 43/50
79/79 [==============================] - 1s 13ms/step - loss: 8.0944 - val_loss: 96.3422
Epoch 44/50
79/79 [==============================] - 1s 12ms/step - loss: 8.4290 - val_loss: 373.7235
Epoch 45/50
79/79 [==============================] - 1s 13ms/step - loss: 9.6937 - val_loss: 190.6326
Epoch 46/50
79/79 [==============================] - 1s 12ms/step - loss: 7.9022 - val_loss: 177.4366
Epoch 47/50
79/79 [==============================] - 1s 12ms/step - loss: 8.1687 - val_loss: 161.2531
Epoch 48/50
79/79 [==============================] - 1s 13ms/step - loss: 9.5011 - val_loss: 140.0809
Epoch 49/50
79/79 [==============================] - 1s 12ms/step - loss: 8.8527 - val_loss: 176.4043
Epoch 50/50
79/79 [==============================] - 1s 12ms/step - loss: 7.7198 - val_loss: 164.9144

四、模型评估

plt.rcParams['font.sans-serif'] = ['SimHei']
plt.rcParams['axes.unicode_minus'] = False
plt.figure(figsize=(5, 3),dpi = 120)

plt.plot(history.history['loss'], label='LSTM Training Loss')
plt.plot(history.history['val_loss'], label='LSTM Validation Loss')

plt.legend()
plt.title('Training and Validation Loss')
plt.show()

在这里插入图片描述

调用模型进行预测

predicted_lstm = model.predict(X_test)

y_test_one = [i[0] for i in y_test]
predicted_y_one = [i[0] for i in predicted_lstm]

plt.figure(figsize=(5, 3),dpi = 120)

plt.plot(y_test_one[:1000], label='真实值')
plt.plot(predicted_y_one[:1000], label='预测值')

plt.legend()
plt.show()

30/30 [==============================] - 0s 3ms/step

在这里插入图片描述

from sklearn import metrics

RMSE_lstm = metrics.mean_squared_error(predicted_lstm,y_test)**0.5
R2_lstm = metrics.r2_score(predicted_lstm,y_test)

RMSE_lstm, R2_lstm

(12.841900251368994, 0.6532699613567474)

完成预测单个时间步达到0.86的R2以后，更改为预测两个时间步，结果如上。

月亮泡面=

关注

16
点赞
踩
12

收藏

觉得还不错? 一键收藏
1
评论
R2 LSTM-火灾温度预测

输入：当前时刻网络输出x_t，上一时刻LSTM输出h_t-1，上一时刻记忆单元c_t-1输出：当前时刻LSTM输出h_t，隐藏状态向量h_t，记忆单元c_t。
复制链接

扫一扫

R2 LSTM-火灾温度预测

LSTM介绍

1. 输入输出

2. 三个门

任务说明：

一、准备工作

二、数据预处理

1. 设置X，y

2. 归一化

3. 划分数据集

三、模型构建，编译，训练

1. 模型构建

2. 模型编译

3. 模型训练

四、模型评估

调用模型进行预测

“相关推荐”对你有帮助么？