Pytorch气温预测实战_拟合插值天气预测-CSDN博客

本文链接：https://blog.csdn.net/m0_46306264/article/details/131997072

数据集

数据有8个特征，一个标签值

自变量	因变量
year	actual:当天的真实最高温度
month
day
week:星期几
temp_1:昨天的最高温度
temp_2:前天的最高温度值
average:在历史中，每年这一天的平均最高温度
friend:朋友猜测的温度

year,month,day,week,temp_2,temp_1,average,actual,friend
2016,1,1,Fri,45,45,45.6,45,29
2016,1,2,Sat,44,45,45.7,44,61
2016,1,3,Sun,45,44,45.8,41,56
2016,1,4,Mon,44,41,45.9,40,53
2016,1,5,Tues,41,40,46,44,41
2016,1,6,Wed,40,44,46.1,51,40
2016,1,7,Thurs,44,51,46.2,45,38
2016,1,8,Fri,51,45,46.3,48,34
2016,1,9,Sat,45,48,46.4,50,47
2016,1,10,Sun,48,50,46.5,52,49
2016,1,11,Mon,50,52,46.7,45,39
2016,1,12,Tues,52,45,46.8,49,61
2016,1,13,Wed,45,49,46.9,55,33
2016,1,14,Thurs,49,55,47,49,58
2016,1,15,Fri,55,49,47.1,48,65
2016,1,16,Sat,49,48,47.3,54,28
2016,1,17,Sun,48,54,47.4,50,47
2016,1,18,Mon,54,50,47.5,54,58
2016,1,19,Tues,50,54,47.6,48,53
2016,1,20,Wed,54,48,47.7,52,61
2016,1,21,Thurs,48,52,47.8,52,57
2016,1,22,Fri,52,52,47.9,57,60
2016,1,23,Sat,52,57,48,48,37
2016,1,24,Sun,57,48,48.1,51,54
2016,1,25,Mon,48,51,48.2,54,63
2016,1,26,Tues,51,54,48.3,56,61
2016,1,27,Wed,54,56,48.4,57,54
2016,1,28,Thurs,56,57,48.4,56,34
2016,1,29,Fri,57,56,48.5,52,49
2016,1,30,Sat,56,52,48.6,48,47
2016,1,31,Sun,52,48,48.7,47,61
2016,2,1,Mon,48,47,48.8,46,51
2016,2,2,Tues,47,46,48.8,51,56
2016,2,3,Wed,46,51,48.9,49,40
2016,2,4,Thurs,51,49,49,49,44
2016,2,5,Fri,49,49,49.1,53,45
2016,2,6,Sat,49,53,49.1,49,56
2016,2,7,Sun,53,49,49.2,51,63
2016,2,8,Mon,49,51,49.3,57,34
2016,2,9,Tues,51,57,49.4,62,57
2016,2,10,Wed,57,62,49.4,56,30
2016,2,11,Thurs,62,56,49.5,55,37
2016,2,12,Fri,56,55,49.6,58,33
2016,2,15,Mon,55,58,49.9,55,53
2016,2,16,Tues,58,55,49.9,56,55
2016,2,17,Wed,55,56,50,57,46
2016,2,18,Thurs,56,57,50.1,53,34
2016,2,19,Fri,57,53,50.2,51,42
2016,2,20,Sat,53,51,50.4,53,43
2016,2,21,Sun,51,53,50.5,51,46
2016,2,22,Mon,53,51,50.6,51,59
2016,2,23,Tues,51,51,50.7,60,43
2016,2,24,Wed,51,60,50.8,59,46
2016,2,25,Thurs,60,59,50.9,61,35
2016,2,26,Fri,59,61,51.1,60,65
2016,2,27,Sat,61,60,51.2,57,61
2016,2,28,Sun,60,57,51.3,53,66
2016,3,1,Tues,53,54,51.5,58,53
2016,3,2,Wed,54,58,51.6,55,37
2016,3,3,Thurs,58,55,51.8,59,71
2016,3,4,Fri,55,59,51.9,57,45
2016,3,5,Sat,59,57,52.1,64,46
2016,3,6,Sun,57,64,52.2,60,49
2016,3,7,Mon,64,60,52.4,53,71
2016,3,8,Tues,60,53,52.5,54,70
2016,3,9,Wed,53,54,52.7,55,57
2016,3,10,Thurs,54,55,52.8,56,50
2016,3,11,Fri,55,56,53,55,36
2016,3,12,Sat,56,55,53.1,52,65
2016,3,13,Sun,55,52,53.3,54,54
2016,3,14,Mon,52,54,53.4,49,44
2016,3,15,Tues,54,49,53.6,51,70
2016,3,16,Wed,49,51,53.7,53,65
2016,3,17,Thurs,51,53,53.9,58,62
2016,3,18,Fri,53,58,54,63,56
2016,3,19,Sat,58,63,54.2,61,62
2016,3,20,Sun,63,61,54.3,55,50
2016,3,21,Mon,61,55,54.5,56,52
2016,3,22,Tues,55,56,54.6,57,64
2016,3,23,Wed,56,57,54.7,53,70
2016,3,24,Thurs,57,53,54.9,54,72
2016,3,25,Fri,53,54,55,57,42
2016,3,26,Sat,54,57,55.2,59,54
2016,3,27,Sun,57,59,55.3,51,39
2016,3,28,Mon,59,51,55.5,56,47
2016,3,29,Tues,51,56,55.6,64,45
2016,3,30,Wed,56,64,55.7,68,57
2016,3,31,Thurs,64,68,55.9,73,56
2016,4,1,Fri,68,73,56,71,41
2016,4,2,Sat,73,71,56.2,63,45
2016,4,3,Sun,71,63,56.3,69,64
2016,4,4,Mon,63,69,56.5,60,45
2016,4,5,Tues,69,60,56.6,57,72
2016,4,6,Wed,60,57,56.8,68,64
2016,4,7,Thurs,57,68,56.9,77,38
2016,4,8,Fri,68,77,57.1,76,41
2016,4,9,Sat,77,76,57.2,66,74
2016,4,10,Sun,76,66,57.4,59,60
2016,4,11,Mon,66,59,57.6,58,40
2016,4,12,Tues,59,58,57.7,60,61
2016,4,13,Wed,58,60,57.9,59,77
2016,4,14,Thurs,60,59,58.1,59,66
2016,4,15,Fri,59,59,58.3,60,40
2016,4,16,Sat,59,60,58.5,68,59
2016,4,17,Sun,60,68,58.6,77,54
2016,4,18,Mon,68,77,58.8,89,39
2016,4,19,Tues,77,89,59,81,61
2016,4,20,Wed,89,81,59.2,81,66
2016,4,21,Thurs,81,81,59.4,73,55
2016,4,22,Fri,81,73,59.7,64,59
2016,4,23,Sat,73,64,59.9,65,57
2016,4,24,Sun,64,65,60.1,55,41
2016,4,25,Mon,65,55,60.3,59,77
2016,4,26,Tues,55,59,60.5,60,75
2016,4,27,Wed,59,60,60.7,61,50
2016,4,28,Thurs,60,61,61,64,73
2016,4,29,Fri,61,64,61.2,61,49
2016,4,30,Sat,64,61,61.4,68,78
2016,5,1,Sun,61,68,61.6,77,75
2016,5,2,Mon,68,77,61.9,87,59
2016,5,3,Tues,77,87,62.1,74,69
2016,5,4,Wed,87,74,62.3,60,61
2016,5,5,Thurs,74,60,62.5,68,56
2016,5,6,Fri,60,68,62.8,77,64
2016,5,7,Sat,68,77,63,82,83
2016,5,8,Sun,77,82,63.2,63,83
2016,5,9,Mon,82,63,63.4,67,64
2016,5,10,Tues,63,67,63.6,75,68
2016,5,11,Wed,67,75,63.8,81,60
2016,5,12,Thurs,75,81,64.1,77,81
2016,5,13,Fri,81,77,64.3,82,67
2016,5,14,Sat,77,82,64.5,65,65
2016,5,15,Sun,82,65,64.7,57,58
2016,5,16,Mon,65,57,64.8,60,53
2016,5,17,Tues,57,60,65,71,55
2016,5,18,Wed,60,71,65.2,64,56
2016,5,19,Thurs,71,64,65.4,63,56
2016,5,20,Fri,64,63,65.6,66,73
2016,5,21,Sat,63,66,65.7,59,49
2016,5,22,Sun,66,59,65.9,66,80
2016,5,23,Mon,59,66,66.1,65,66
2016,5,24,Tues,66,65,66.2,66,67
2016,5,25,Wed,65,66,66.4,66,60
2016,5,26,Thurs,66,66,66.5,65,85
2016,5,27,Fri,66,65,66.7,64,73
2016,5,28,Sat,65,64,66.8,64,64
2016,5,29,Sun,64,64,67,64,76
2016,5,30,Mon,64,64,67.1,71,69
2016,5,31,Tues,64,71,67.3,79,85
2016,6,1,Wed,71,79,67.4,75,58
2016,6,2,Thurs,79,75,67.6,71,77
2016,6,3,Fri,75,71,67.7,80,55
2016,6,4,Sat,71,80,67.9,81,76
2016,6,5,Sun,80,81,68,92,54
2016,6,6,Mon,81,92,68.2,86,71
2016,6,7,Tues,92,86,68.3,85,58
2016,6,8,Wed,86,85,68.5,67,81
2016,6,9,Thurs,85,67,68.6,65,80
2016,6,10,Fri,67,65,68.8,67,73
2016,6,11,Sat,65,67,69,65,87
2016,6,12,Sun,67,65,69.1,70,83
2016,6,13,Mon,65,70,69.3,66,79
2016,6,14,Tues,70,66,69.5,60,85
2016,6,15,Wed,66,60,69.7,67,69
2016,6,16,Thurs,60,67,69.8,71,87
2016,6,17,Fri,67,71,70,67,54
2016,6,18,Sat,71,67,70.2,65,77
2016,6,19,Sun,67,65,70.4,70,58
2016,6,20,Mon,65,70,70.6,76,79
2016,6,21,Tues,70,76,70.8,73,57
2016,6,22,Wed,76,73,71,75,78
2016,6,23,Thurs,73,75,71.3,68,56
2016,6,24,Fri,75,68,71.5,69,65
2016,6,25,Sat,68,69,71.7,71,89
2016,6,26,Sun,69,71,71.9,78,70
2016,6,27,Mon,71,78,72.2,85,84
2016,6,28,Tues,78,85,72.4,79,67
2016,6,29,Wed,85,79,72.6,74,81
2016,6,30,Thurs,79,74,72.8,73,87
2016,7,1,Fri,74,73,73.1,76,93
2016,7,2,Sat,73,76,73.3,76,84
2016,7,3,Sun,76,76,73.5,71,85
2016,7,4,Mon,76,71,73.8,68,86
2016,7,5,Tues,71,68,74,69,62
2016,7,6,Wed,68,69,74.2,76,86
2016,7,7,Thurs,69,76,74.4,68,72
2016,7,8,Fri,76,68,74.6,74,77
2016,7,9,Sat,68,74,74.9,71,60
2016,7,10,Sun,74,71,75.1,74,95
2016,7,11,Mon,71,74,75.3,74,71
2016,7,12,Tues,74,74,75.4,77,71
2016,7,13,Wed,74,77,75.6,75,56
2016,7,14,Thurs,77,75,75.8,77,77
2016,7,15,Fri,75,77,76,76,75
2016,7,16,Sat,77,76,76.1,72,61
2016,7,17,Sun,76,72,76.3,80,88
2016,7,18,Mon,72,80,76.4,73,66
2016,7,19,Tues,80,73,76.6,78,90
2016,7,20,Wed,73,78,76.7,82,66
2016,7,21,Thurs,78,82,76.8,81,84
2016,7,22,Fri,82,81,76.9,71,70
2016,7,23,Sat,81,71,77,75,86
2016,7,24,Sun,71,75,77.1,80,75
2016,7,25,Mon,75,80,77.1,85,81
2016,7,26,Tues,80,85,77.2,79,74
2016,7,27,Wed,85,79,77.3,83,79
2016,7,28,Thurs,79,83,77.3,85,76
2016,7,29,Fri,83,85,77.3,88,77
2016,7,30,Sat,85,88,77.3,76,70
2016,7,31,Sun,88,76,77.4,73,95
2016,8,1,Mon,76,73,77.4,77,65
2016,8,2,Tues,73,77,77.4,73,62
2016,8,3,Wed,77,73,77.3,75,93
2016,8,4,Thurs,73,75,77.3,80,66
2016,8,5,Fri,75,80,77.3,79,71
2016,8,6,Sat,80,79,77.2,72,60
2016,8,7,Sun,79,72,77.2,72,95
2016,8,8,Mon,72,72,77.1,73,65
2016,8,9,Tues,72,73,77.1,72,94
2016,8,10,Wed,73,72,77,76,68
2016,8,11,Thurs,72,76,76.9,80,80
2016,8,12,Fri,76,80,76.9,87,81
2016,8,13,Sat,80,87,76.8,90,73
2016,8,14,Sun,87,90,76.7,83,65
2016,8,15,Mon,90,83,76.6,84,70
2016,8,16,Tues,83,84,76.5,81,90
2016,8,23,Tues,84,81,75.7,79,89
2016,8,28,Sun,81,79,75,75,85
2016,8,30,Tues,79,75,74.6,70,63
2016,9,3,Sat,75,70,73.9,67,68
2016,9,4,Sun,70,67,73.7,68,64
2016,9,5,Mon,67,68,73.5,68,54
2016,9,6,Tues,68,68,73.3,68,79
2016,9,7,Wed,68,68,73,67,70
2016,9,8,Thurs,68,67,72.8,72,56
2016,9,9,Fri,67,72,72.6,74,78
2016,9,10,Sat,72,74,72.3,77,91
2016,9,11,Sun,74,77,72.1,70,70
2016,9,12,Mon,77,70,71.8,74,90
2016,9,13,Tues,70,74,71.5,75,82
2016,9,14,Wed,74,75,71.2,79,77
2016,9,15,Thurs,75,79,71,71,64
2016,9,16,Fri,79,71,70.7,75,52
2016,9,17,Sat,71,75,70.3,68,84
2016,9,18,Sun,75,68,70,69,90
2016,9,19,Mon,68,69,69.7,71,88
2016,9,20,Tues,69,71,69.4,67,81
2016,9,21,Wed,71,67,69,68,76
2016,9,22,Thurs,67,68,68.7,67,56
2016,9,23,Fri,68,67,68.3,64,61
2016,9,24,Sat,67,64,68,67,64
2016,9,25,Sun,64,67,67.6,76,62
2016,9,26,Mon,67,76,67.2,77,74
2016,9,27,Tues,76,77,66.8,69,64
2016,9,28,Wed,77,69,66.5,68,62
2016,9,29,Thurs,69,68,66.1,66,57
2016,9,30,Fri,68,66,65.7,67,74
2016,10,1,Sat,66,67,65.3,63,54
2016,10,2,Sun,67,63,64.9,65,82
2016,10,3,Mon,63,65,64.5,61,49
2016,10,4,Tues,65,61,64.1,63,60
2016,10,5,Wed,61,63,63.7,66,48
2016,10,6,Thurs,63,66,63.3,63,55
2016,10,7,Fri,66,63,62.9,64,78
2016,10,8,Sat,63,64,62.5,68,73
2016,10,9,Sun,64,68,62.1,57,55
2016,10,10,Mon,68,57,61.8,60,62
2016,10,11,Tues,57,60,61.4,62,58
2016,10,12,Wed,60,62,61,66,52
2016,10,13,Thurs,62,66,60.6,60,57
2016,10,14,Fri,66,60,60.2,60,78
2016,10,15,Sat,60,60,59.9,62,46
2016,10,16,Sun,60,62,59.5,60,40
2016,10,17,Mon,62,60,59.1,60,62
2016,10,18,Tues,60,60,58.8,61,53
2016,10,19,Wed,60,61,58.4,58,41
2016,10,20,Thurs,61,58,58.1,62,43
2016,10,21,Fri,58,62,57.8,59,44
2016,10,22,Sat,62,59,57.4,62,44
2016,10,23,Sun,59,62,57.1,62,67
2016,10,24,Mon,62,62,56.8,61,70
2016,10,25,Tues,62,61,56.5,65,70
2016,10,26,Wed,61,65,56.2,58,41
2016,10,27,Thurs,65,58,55.9,60,39
2016,10,28,Fri,58,60,55.6,65,52
2016,10,29,Sat,60,65,55.3,68,65
2016,10,31,Mon,65,68,54.8,59,62
2016,11,1,Tues,68,59,54.5,57,61
2016,11,2,Wed,59,57,54.2,57,70
2016,11,3,Thurs,57,57,53.9,65,35
2016,11,4,Fri,57,65,53.7,65,38
2016,11,5,Sat,65,65,53.4,58,41
2016,11,6,Sun,65,58,53.2,61,71
2016,11,7,Mon,58,61,52.9,63,35
2016,11,8,Tues,61,63,52.7,71,49
2016,11,9,Wed,63,71,52.4,65,42
2016,11,10,Thurs,71,65,52.2,64,38
2016,11,11,Fri,65,64,51.9,63,55
2016,11,12,Sat,64,63,51.7,59,63
2016,11,13,Sun,63,59,51.4,55,64
2016,11,14,Mon,59,55,51.2,57,42
2016,11,15,Tues,55,57,51,55,46
2016,11,16,Wed,57,55,50.7,50,34
2016,11,17,Thurs,55,50,50.5,52,57
2016,11,18,Fri,50,52,50.3,55,35
2016,11,19,Sat,52,55,50,57,56
2016,11,20,Sun,55,57,49.8,55,30
2016,11,21,Mon,57,55,49.5,54,67
2016,11,22,Tues,55,54,49.3,54,58
2016,11,23,Wed,54,54,49.1,49,38
2016,11,24,Thurs,54,49,48.9,52,29
2016,11,25,Fri,49,52,48.6,52,41
2016,11,26,Sat,52,52,48.4,53,58
2016,11,27,Sun,52,53,48.2,48,53
2016,11,28,Mon,53,48,48,52,44
2016,11,29,Tues,48,52,47.8,52,50
2016,11,30,Wed,52,52,47.6,52,44
2016,12,1,Thurs,52,52,47.4,46,39
2016,12,2,Fri,52,46,47.2,50,41
2016,12,3,Sat,46,50,47,49,58
2016,12,4,Sun,50,49,46.8,46,53
2016,12,5,Mon,49,46,46.6,40,65
2016,12,6,Tues,46,40,46.4,42,56
2016,12,7,Wed,40,42,46.3,40,62
2016,12,8,Thurs,42,40,46.1,41,36
2016,12,9,Fri,40,41,46,36,54
2016,12,10,Sat,41,36,45.9,44,65
2016,12,11,Sun,36,44,45.7,44,35
2016,12,12,Mon,44,44,45.6,43,42
2016,12,13,Tues,44,43,45.5,40,46
2016,12,14,Wed,43,40,45.4,39,49
2016,12,15,Thurs,40,39,45.3,39,46
2016,12,16,Fri,39,39,45.3,35,39
2016,12,17,Sat,39,35,45.2,35,38
2016,12,18,Sun,35,35,45.2,39,36
2016,12,19,Mon,35,39,45.1,46,51
2016,12,20,Tues,39,46,45.1,51,62
2016,12,21,Wed,46,51,45.1,49,39
2016,12,22,Thurs,51,49,45.1,45,38
2016,12,23,Fri,49,45,45.1,40,35
2016,12,24,Sat,45,40,45.1,41,39
2016,12,25,Sun,40,41,45.1,42,31
2016,12,26,Mon,41,42,45.2,42,58
2016,12,27,Tues,42,42,45.2,47,47
2016,12,28,Wed,42,47,45.3,48,58
2016,12,29,Thurs,47,48,45.3,48,65
2016,12,30,Fri,48,48,45.4,57,42
2016,12,31,Sat,48,57,45.5,40,57

数据集处理

读取数据

# -*-coding:utf-8-*-
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import torch
import torch.optim as optim
import warnings
warnings.filterwarnings("ignore")
#   year  month  day  week  temp_2  temp_1  average  actual(y)  friend
features=pd.read_csv("temps.csv")

# print(features.head())
#数据维度 (348, 9)
print("数据维度",features.shape)
print(features.head(5))

数据维度 (348, 9)
year month day week temp_2 temp_1 average actual friend
0 2016 1 1 Fri 45 45 45.6 45 29
1 2016 1 2 Sat 44 45 45.7 44 61
2 2016 1 3 Sun 45 44 45.8 41 56
3 2016 1 4 Mon 44 41 45.9 40 53
4 2016 1 5 Tues 41 40 46.0 44 41

把年月日转换为datetime格式

#处理事件数据
import datetime
#分别得到年、月、日
years=features["year"]
months=features["month"]
days=features["day"]

#datetime格式
"""
['2016-1-1', '2016-1-2', '2016-1-3', '2016-1-4', '2016-1-5']
"""
dates=[str(int(year))+"-"+str(int(month))+"-"+str(int(day)) for year,month,day in zip(years,months,days)]
print(dates[:5])


"""
[datetime.datetime(2016, 1, 1, 0, 0), 
datetime.datetime(2016, 1, 2, 0, 0), 
datetime.datetime(2016, 1, 3, 0, 0), 
datetime.datetime(2016, 1, 4, 0, 0), 
datetime.datetime(2016, 1, 5, 0, 0)]
"""
dates=[datetime.datetime.strptime(date, "%Y-%m-%d") for date in dates]
print(dates[:5])

['2016-1-1', '2016-1-2', '2016-1-3', '2016-1-4', '2016-1-5']
[datetime.datetime(2016, 1, 1, 0, 0),

datetime.datetime(2016, 1, 2, 0, 0),

datetime.datetime(2016, 1, 3, 0, 0),

datetime.datetime(2016, 1, 4, 0, 0),

datetime.datetime(2016, 1, 5, 0, 0)]

画图显示数据

plt.subplots(nrows=2, ncols=2, figsize = (20,20)) 2行，2列，图像大小20*20
fig.autofmt_xdate(rotation = 45) X轴的字体旋转角度，也就是dates所对应的年月日信息显示旋转角度，这里是45°，如：2016-01等，倾斜显示

ax1.plot(dates, dataset['actual']) X轴为dates(年月日)，Y轴为actual(实际的真实温度值)
ax1.set_xlabel(''); ax1.set_ylabel('Temperature'); ax1.set_title('Actual Max Temp') X轴标签为空，Y轴标签为Temperature，整体标题为Actual Max Temp
其他的同理

plt.tight_layout(pad=1, h_pad=1, w_pad=1) #子图间隔有多大
pad：图形边和子图的边之间进行填充
h_pad，w_pad：相邻子图的边之间的填充(高度/宽度)

def drawData(dates,features):
    #准备画图
    #指定默认风格
    plt.style.use("fivethirtyeight")

    #设置布局
    fig, ((ax1,ax2),(ax3,ax4)) = plt.subplots(nrows=2,ncols=2,figsize=(10,10))
    #X轴标签倾斜显示
    fig.autofmt_xdate(rotation=45)

    #标签值
    ax1.plot(dates,features["actual"])
    ax1.set_xlabel(" "); ax1.set_ylabel("Actual Temperature"); ax1.set_title("Max Temp")

    #昨天
    ax2.plot(dates,features["temp_1"])
    ax2.set_xlabel(" "); ax2.set_ylabel("temp_1 Temperature"); ax2.set_title("Previous Max Temp")

    #前天
    ax3.plot(dates,features["temp_2"])
    ax3.set_xlabel("Date"); ax3.set_ylabel("temp_2 Temperatiure"); ax3.set_title("Two Days Prior Previous Max Temp")

    ax4.plot(dates,features["friend"])
    ax4.set_xlabel("Date"); ax4.set_ylabel("friend Temperatiure"); ax4.set_title("Friend Estimate")

    #子图间隔多大
    plt.tight_layout(pad=2)
    plt.show()

drawData(dates,features)

把星期几转为独热编码

#读热编码
#year  month  day  temp_2  ... week_Fri week_Mon week_Sat week_Sun  week_Thurs  week_Tues  week_Wed
features=pd.get_dummies(features)
print( features.head(8) )

year month day temp_2 ... week_Sun week_Thurs week_Tues week_Wed
0 2016 1 1 45 ... 0 0 0 0
1 2016 1 2 44 ... 0 0 0 0
2 2016 1 3 45 ... 1 0 0 0
3 2016 1 4 44 ... 0 0 0 0
4 2016 1 5 41 ... 0 0 1 0
5 2016 1 6 40 ... 0 0 0 1
6 2016 1 7 44 ... 0 1 0 0
7 2016 1 8 51 ... 0 0 0 0

提取actual的实际值

#标签，把标签值actual提取出来
#labels=[45 44 41 40 44 51 45 48 50 ....]
labels=np.array(features["actual"])

#在特征中去掉标签, 去掉actual，因为actual是y
features=features.drop("actual",axis=1)

#名字单独保存一下，以备后患
feature_list=list(features.columns)
print(feature_list)

['year', 'month', 'day', 'temp_2', 'temp_1', 'average', 'friend', 'week_Fri', 'week_Mon', 'week_Sat', 'week_Sun', 'week_Thurs', 'week_Tues', 'week_Wed']

标准化数据

(1)把features变成了数值，用科学计数法,(348,14)

features=np.array(features)
print(features)

[[2.016e+03 1.000e+00 1.000e+00 ... 0.000e+00 0.000e+00 0.000e+00]
[2.016e+03 1.000e+00 2.000e+00 ... 0.000e+00 0.000e+00 0.000e+00]
[2.016e+03 1.000e+00 3.000e+00 ... 0.000e+00 0.000e+00 0.000e+00]
...
[2.016e+03 1.200e+01 2.900e+01 ... 1.000e+00 0.000e+00 0.000e+00]
[2.016e+03 1.200e+01 3.000e+01 ... 0.000e+00 0.000e+00 0.000e+00]
[2.016e+03 1.200e+01 3.100e+01 ... 0.000e+00 0.000e+00 0.000e+00]]

(2)标准化 (x-平均值)/标准差

#标准化 (x-平均值)/标准差
from sklearn import  preprocessing
input_features=preprocessing.StandardScaler().fit_transform(features)

print( input_features )

[[ 0. -1.5678393 -1.65682171 ... -0.40482045 -0.41913682
-0.40482045]
[ 0. -1.5678393 -1.54267126 ... -0.40482045 -0.41913682
-0.40482045]
[ 0. -1.5678393 -1.4285208 ... -0.40482045 -0.41913682
-0.40482045]
...
[ 0. 1.5810006 1.53939107 ... 2.47023092 -0.41913682
-0.40482045]
[ 0. 1.5810006 1.65354153 ... -0.40482045 -0.41913682
-0.40482045]
[ 0. 1.5810006 1.76769198 ... -0.40482045 -0.41913682
-0.40482045]]

第一种方式构建网络模型

(1)构建x和y矩阵。

构建tensor类型的x和y变量

#x:torch.Size([348, 14])
x=torch.tensor(input_features,dtype=float)
y=torch.tensor(labels,dtype=float)

(2)权重参数初始化

#权重参数初始化
weights=torch.randn((14,128),dtype=float,requires_grad=True)
biases=torch.randn(128,dtype=float ,requires_grad=True)
weights2=torch.randn((128,1),dtype=float,requires_grad=True)
biases2=torch.randn(1,dtype=float ,requires_grad=True)
#用于梯度下降的学习率
learning_rate=0.001
#记录损失值
losses=[]

(3)搭建网络模型

#搭建网络模型
for i in range(1000):
    #计算隐层
    hidden=x.mm(weights)+biases
    #激活函数
    hidden=torch.relu(hidden)
    #预测结果
    predictions=hidden.mm(weights2)+biases2

    #计算损失
    #cost/w 1/2n
    loss=torch.mean((predictions - y)**2)
    losses.append(loss.data.numpy())
    #打印损失值
    if i%100==0:
        print("loss:",loss)

    #反向传播计算
    loss.backward()

    #更新参数
    weights.data.add_(- learning_rate*weights.grad.data)
    biases.data.add_(- learning_rate*biases.grad.data)
    weights2.data.add_(- learning_rate*weights2.grad.data)
    biases2.data.add_(- learning_rate*biases2.grad.data)

    #每次迭代都得清空，梯度不清零会累加的
    weights.grad.data.zero_()
    biases.grad.data.zero_()
    weights2.grad.data.zero_()
    biases2.grad.data.zero_()

print(predictions.shape)

loss: tensor(1214.2573, dtype=torch.float64, grad_fn=<MeanBackward0>)
loss: tensor(155.9543, dtype=torch.float64, grad_fn=<MeanBackward0>)
loss: tensor(147.3351, dtype=torch.float64, grad_fn=<MeanBackward0>)
loss: tensor(144.7511, dtype=torch.float64, grad_fn=<MeanBackward0>)
loss: tensor(143.4206, dtype=torch.float64, grad_fn=<MeanBackward0>)
loss: tensor(142.5542, dtype=torch.float64, grad_fn=<MeanBackward0>)
loss: tensor(141.9567, dtype=torch.float64, grad_fn=<MeanBackward0>)
loss: tensor(141.5120, dtype=torch.float64, grad_fn=<MeanBackward0>)
loss: tensor(141.1755, dtype=torch.float64, grad_fn=<MeanBackward0>)
loss: tensor(140.9053, dtype=torch.float64, grad_fn=<MeanBackward0>)
torch.Size([348, 1])

(4) 把actual的实际值和prediction画在一张图

loss是140.9，所以拟合效果非常差

第二种方式构建神经网络

(1)数据预处理

# -*-coding:utf-8-*-
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import torch
import torch.optim as optim
import warnings
warnings.filterwarnings("ignore")
#   year  month  day  week  temp_2  temp_1  average  actual(y)  friend
features=pd.read_csv("temps.csv")

# print(features.head())
#数据维度 (348, 9)
print("数据维度",features.shape)


#处理事件数据
import datetime
#分别得到年、月、日
years=features["year"]
months=features["month"]
days=features["day"]

#datetime格式
dates=[str(int(year))+"-"+str(int(month))+"-"+str(int(day)) for year,month,day in zip(years,months,days)]
dates=[datetime.datetime.strptime(date, "%Y-%m-%d") for date in dates]

print(dates[:5])


#hotcode
#year  month  day  temp_2  ... week_Fri week_Mon week_Sat week_Sun  week_Thurs  week_Tues  week_Wed
features=pd.get_dummies(features)
# print( features.head(5) )

#标签
labels=np.array(features["actual"])

#在特征中去掉标签, 去掉actual，因为actual是y
features=features.drop("actual",axis=1)

#名字单独保存一下，以备后患
feature_list=list(features.columns)

#变成了数值，用科学计数法,(348,14)
features=np.array(features)
# print(features)

#标准化 (x-平均值)/标准差
from sklearn import  preprocessing
input_features=preprocessing.StandardScaler().fit_transform(features)

# print( input_features )
# print( input_features[0] )

#构建网络模型
#x:torch.Size([348, 14])
x=torch.tensor(input_features,dtype=float)
y=torch.tensor(labels,dtype=float)

(2)构建模型，计算预测值

#更简单的构建网络模型
#input_size:特征数:14
input_size=input_features.shape[1]

#隐藏层有128个神经元
hidden_size=128
#神经网络的最终输出1个数，即predictions_actual
output_size=1
#每次读取16行数据
batch_size=16

#Sequential:按顺序执行
my_nn=torch.nn.Sequential(
    #第一层神经元(14,128)
    torch.nn.Linear(input_size,hidden_size),
    torch.nn.Sigmoid(),
    #第二层神经元(128,1)
    torch.nn.Linear(hidden_size,output_size)
)
#损失函数
cost=torch.nn.MSELoss(reduction="mean")

#梯度下降，更新权重参数
#Adam:很强的工具，
optimizer=torch.optim.Adam(my_nn.parameters(),lr=0.001)

#训练网络
losses=[]
for i in range (2000):
    batch_loss=[]
    #MINI-Batch方法进行训练,0~348,每次训练选batch_size(16)个样本
    for start in range (0,len(input_features),batch_size):
        end=start+batch_size if start+batch_size <len(input_features) else len(input_features)
        xx=torch.tensor(input_features[start:end],dtype=torch.float,requires_grad=True)
        yy=torch.tensor(labels[start:end],dtype=torch.float,requires_grad=True)
        prediction=my_nn(xx)
        loss=cost(prediction,yy)
        #梯度清零
        optimizer.zero_grad()
        #反向传播
        loss.backward(retain_graph=True)
        #梯度下降：step
        optimizer.step()
        batch_loss.append(loss.data.numpy())

    #计算损失
    if i%100==0:
        losses.append(np.mean(batch_loss))
        print(i,np.mean(batch_loss))


#预测训练结果
x=torch.tensor(input_features,dtype=torch.float)
predict=my_nn(x).data.numpy()

0 3938.2195
100 37.869488
200 35.65198
300 35.276585
400 35.106327
500 34.972412
600 34.85649
700 34.742622
800 34.62497
900 34.503555

(3) 画图

#转换日期格式
#datetime格式
dates=[str(int(year))+"-"+str(int(month))+"-"+str(int(day)) for year,month,day in zip(years,months,days)]
dates=[datetime.datetime.strptime(date, "%Y-%m-%d") for date in dates]

#true创建
true_data=pd.DataFrame(data={"date":dates,"actual":labels})

#同理，再创建一个来存日期和其对应的模型预测值
months=features[:,feature_list.index("month")]
days=features[:,feature_list.index("day")]
years=features[:,feature_list.index("year")]

test_dates=[str(int(year))+"-"+str(int(month))+"-"+str(int(day)) for year,month,day in zip(years,months,days)]
#          [datetime.datetime.strptime(date, "%Y-%m-%d") for date in dates]
test_dates=[datetime.datetime.strptime(date, "%Y-%m-%d") for date in test_dates]

predictions_data=pd.DataFrame(data={"date":test_dates,"prediction":predict.reshape(-1)})


#真实值
plt.plot(true_data["date"],true_data["actual"],"b-",label="actual")

#预测值
plt.plot(predictions_data["date"],predictions_data["prediction"],"ro",label="prediction")
plt.xticks(rotation = 60);
# plt.xticks(rotation="60")
plt.legend()
#图名
plt.xlabel("Date");plt.ylabel("Maximum Temperature(F)");plt.title("Actual and Predicted Values");
plt.show()