Python 实现长短记忆神经⽹络（LSTM ）预测极限学习机（ELM ）残差并累加的
时间序列预测
本实验使⽤环境为Anaconda3 Jupyter，调⽤Sklearn包，调⽤keras包，请提前准备好。
1.引⼊⼀些常见包
主要有包、numpy包、metrics包、pandas包等。
2.引⼊数据
just a friend其中data为全部数据、traffic_feature为特征集、traffic_target⽬标集
对数据进⾏分割，本⽂使⽤70%为分割点。import csv import numpy as np import time from sklearn .preprocessing import StandardScaler from sklearn .model_lection import train_test_split from sklearn .enmble import RandomForestClassifier from sklearn .metrics import accuracy_score from sklearn .metrics import confusion_matrix from sklearn .metrics import classification_report from sklearn .metrics import explained_variance_score from sklearn import metrics from sklearn .svm import SVR import matplotlib .pyplot as plt from pandas import DataFrame from pandas import Se
ries from pandas import concat from pandas import read_csv from pandas import datetime from sklearn .metrics import mean_squared_error from sklearn .preprocessing import MinMaxScaler from keras .models import Sequential from keras .layers import Den from keras .layers import LSTM from math import sqrt from matplotlib import pyplot import numpy
1
2
3
4
5
即时通信软件6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26data =[]traffic_feature =[]traffic_target =[]csv_file = csv .reader (open ('GoodData.csv'))for content in csv_file : content =list (map (float ,content )) if len (content )!=0: data .append (content ) traffic_feature .append (content [0:4]) traffic_target .append (content [-1]) traffic_feature =np .array (traffic_feature )traffic_target =np .array (traffic_target )data =np .array (data )
1
2
3
4
5
6
7
8
9
10
读题
11
12
13
14
15
对后30%的⽬标值继续分割，分割点仍然为70%，预留做对照。
3.数据标准化
使⽤StandardScaler()⽅法将特征数据标准化归⼀化。
4.使⽤ELM 算法预测feature_train =traffic_feature [0:int (len (traffic_feature )*0.7)]feature_test =traffic_feature [int (len (traffic_feature )*0.7):int (len (traffic_feature ))]target_train =traffic_target [0:int (len (traffic_target )*0.7)]target_test =traffic_target [int (len (traffic_target )*0.7):int (len (traffic_target ))]1
2
3
4
5target_test_last =target_test [int (len (target_test )*0.7):int (len (target_test ))]
1scaler = StandardScaler () # 标准化转换scaler .fit (traffic_feature ) # 训练标准化对象traffic_feature = scaler .transform (traffic_feature ) # 转换数据集
1
2
3
class HiddenLayer : def __init__(lf , x , num ): # x ：输⼊矩阵 num ：隐含层神经元个数 row = x .shape [0] columns = x .shape [1] rnd = np .random .RandomState (9999) lf .w = rnd .uniform (-1, 1, (columns , num )) # lf .b = np .zeros ([row , num ], dtype =float ) # 随机设定隐含层神经元阈值，即bi 的值 for i in range (num ): rand_b = rnd .uniform (-0.4, 0.4) # 随机产⽣-0.4 到 0.4 之间的数 for j in range (row ): lf .b [j , i ] = rand_b # 设定输⼊层与隐含层的连接权值 lf .h = lf .sigmoid (np .dot (x , lf .w ) + lf .b ) # 计算隐含层输出矩阵H lf .H_ = np .linalg .pinv (lf .h ) # 获取H 的逆矩阵 # print(lf.H_.shape) # 定义激活函数g(x) ，需为⽆限可微函数 def sigmoid (lf , x ): print (x ) return 1.0 / (1 +
np .exp (-x )) ''' 若进⾏分析的训练数据为回归问题，则使⽤该⽅式 ，计算隐含层输出权值，即β ''' def regressor_train (lf , T ): C = 2 I = len (T ) sub_former = np .dot (np .transpo (lf .h ), lf .h ) + I / C all_m = np .dot (np .linalg .pinv (sub_former ), np .transpo (lf .h )) T = T .reshape (-1, 1) lf .beta = np .dot (all_m , T ) return lf .beta """ 计算隐含层输出权值，即β """ def classifisor_train (lf , T ): en_one = OneHotEncoder () # print(T) T = en_one .fit_transform (T .reshape (-1, 1)).toarray () # 独热编码之后⼀定要⽤toarray()转换成正常的数组 # print(T) C = 3 I = len (T ) sub_former = np .dot (np .transpo (lf .h ), lf .h ) + I / C all_m = np .dot (np .linalg .pinv (sub_former ), np .transpo (lf .h )) lf .beta = np .dot (all_m , T ) return lf .beta def regressor_test (lf , test_x ): b_row = test_x .shape [0] h = lf .sigmoid (np .dot (test_x , lf .w ) + lf .b [:b_row , :]) result = np .dot (h , lf .beta ) return result def classifisor_test (lf , test_x ): b_row = test_x .shape [0] h = lf .sigmoid (np .dot (test_x , lf .w ) + lf .b [:b_row , :]) result = np .dot (h , lf .beta ) result = [item .tolist ().index (max (item .tolist ())) for item in result ] return result
1
2
会议纪要英文
3
4
5
6竖琴螺
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
生活压力大怎么办24
25
26
27
28
29
进度英语30
31
rossignol
32
33
34
35
36
37
38
39
40
41
42
43
44the graduate
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
跑起来～
设置神经元个数为8，可以⾃⾏调优。
rosario
结果如下：
EVS等于0.8705
5.使⽤真实值减预测值，获的ELM 的残差import matplotlib .pyplot as plt from sklearn .metrics import explained_variance_score a = HiddenLayer (feature_train ,8)a .regressor_train (target_train )result = a .regressor_test (feature_test )plt .plot (result )#测试数组plt .plot (target_test )#测试数组plt .legend (['ELM','TRUE'])fig = plt .gcf ()fig .t_size_inches (18.5, 10.5)plt .title ("ELM") # 标题plt .show ()print ("EVS:",explained_variance_score (target_test ,result ))1
2
3
4
5
6
7
8
9
10
11
12
13
残差值变化如下：
将预测的后30%截取，预留做对照。
6.对残差值使⽤长短记忆（LSTM ）预测
使⽤残差值的前70%作为测试集，使⽤后30%作为验证集。a =[]#真实值for i in target_test : a .append (i )b =[]#预测值for i in result : b .append (i [0])c =[]#残差值num =[]for inx ,i in enumerate (a ): c .append (b [inx ]-i ) num .append (inx )plt .plot (c )#残差fig = plt .gcf ()fig .t_size_inches (18.5,5)plt .xlim (0,1560)plt .title ("Residual Signal") # 标题plt .show ()
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18result_last =b [int (len (b )*0.7):int (len (b ))]
1
