在日期时间中每 10 分钟绘制一次
Posted
技术标签:
【中文标题】在日期时间中每 10 分钟绘制一次【英文标题】：Plot for every 10 minutes in datetime 【发布时间】：2020-02-15 13:49:51 【问题描述】：我使用的“df”对于每个datetime 都有多行。我想每 10 分钟绘制一个具有相同 datetime 的所有坐标的散点图。每个位置都有一个数据条目，每 10 分钟在df_data
如果我手动将时间输入t_list = [datetime(2017, 12, 23, 06, 00, 00), datetime(2017, 12, 23, 06, 10, 00), datetime(2017, 12, 23, 06, 20, 00)]，它会起作用，但我想用使用df 中的日期的东西来替换它，这样我就可以将它用于多个数据集。
import pandas as pd
import matplotlib.pyplot as plt
from datetime import datetime, timedelta
import numpy as np
df_data = pd.read_csv('C:\data.csv')
df_data['datetime'] = pd.to_datetime(df_data['TimeStamp'] )
df = df_data[(df_data['datetime']>= datetime(2017, 12, 23, 06,00, 00)) &
(df_data['datetime']< datetime(2017, 12, 23, 07, 00, 00))]
##want a time array for all of the datetimes in the df
t_list = [datetime(2017, 12, 23, 06, 00, 00), datetime(2017, 12, 23, 06, 10, 00),
datetime(2017, 12,
23, 06, 20, 00)]
for t in t_list:
t_end = t + timedelta(minutes = 10)
t_text = t.strftime("%d-%b-%Y (%H:%M)")
#boolean indexing with multiple conditions, you should wrap each single condition in brackets
df_t = df[(df['datetime']>=t) & (df['datetime']<t_end)]
#get data into variable
ws = df_t['Sp_mean']
lat = df_t['x']
lon = df_t['y']
col = 0.75
#calc min/max for setting scale on images
min_ws = df['Sp_mean'].min()
max_ws = df['Sp_mean'].max()
plt.figure(figsize=(15,10))
plt.scatter(lon, lat, c=ws,s=300, vmin=min_ws, vmax=max_ws)
plt.title('event' + t_text,fontweight = 'bold',fontsize=18)
plt.show()
我尝试了几种方法来尝试将datetime 复制为可迭代列表，但没有给我想要的结果，最近的结果如下：
date_arrray = np.arange(np.datetime64(df['datetime']))
df['timedelta'] = pd.to_timedelta(df['datetime'])
示例数据集
【问题讨论】：
您能否发布数据集的样本或至少发布您使用的日期格式？ 您是否要进行“分组”？ -- pandas.pydata.org/pandas-docs/stable/reference/api/… @gustavovelascoh - 编辑问题以包含数据集的 sn-p @squar_o 我是否正确理解您希望每 10 分钟间隔一个散点图？t_list = df['TimeStamp'].unique()。然后对其进行排序并迭代值以获得正确的切片。
【参考方案1】：
熊猫你不熟悉的接缝。您应该检查 resample function。
让df_data 成为您的原始数据：
# make a DatetimeIndex and resample it to 10-Min interval
df_data.index = pd.to_datetime(df_data['TimeStamp'])
resampled_data = df_data.resample('10Min')
# loop it:
min_ws = df['Sp_mean'].min()
max_ws = df['Sp_mean'].max()
col = 0.75
for start_time, sampled_df in resampled_data:
ws = sampled_df['Sp_mean']
lat = sampled_df['x']
lon = sampled_df['y']
plt.figure(figsize=(15,10))
plt.scatter(lon, lat, c=ws,s=300, vmin=min_ws, vmax=max_ws)
plt.title('event' + start_time.strftime('%Y-%m-%d %H:%M:%S'),fontweight = 'bold',fontsize=18)
plt.show()
【讨论】：
我应该明确表示我的数据已经每 10 分钟采样一次 - 进行了编辑以反映这一点。 @squar_o 这不会改变任何事情。您仍然可以使用resample 进行分组和统计。【参考方案2】：
我试过这个数据集：
datetime x y
0 31/10/2017 23:50 1 9
1 31/10/2017 23:50 1 9
2 31/10/2017 23:50 1 9
3 31/10/2017 23:40 1 9
4 31/10/2017 23:40 1 9
5 31/10/2017 23:40 1 9
6 31/10/2017 23:30 1 9
7 31/10/2017 23:30 1 9
8 31/10/2017 23:20 1 9
还有这段代码：
a = [["31/10/2017 23:50", 1,9],["31/10/2017 23:50", 1,9],["31/10/2017 23:50", 1,9],["31/10/2017 23:40", 1,9],["31/10/2017 23:40", 1,9],["31/10/2017 23:40", 1,9],["31/10/2017 23:30", 1,9],["31/10/2017 23:30", 1,9],["31/10/2017 23:20", 1,9]]
df = pd.DataFrame(a,columns=["TimeStamp","x","y"])
df["datetime"] = pd.to_datetime(df["TimeStamp"])
t_list = df.groupby("datetime").all().index
print(t_list)
# DatetimeIndex(['2017-10-31 23:20:00', '2017-10-31 23:30:00',
# ... '2017-10-31 23:40:00', '2017-10-31 23:50:00'],
# ... dtype='datetime64[ns]', name='datetime', freq=None)
【讨论】：
【参考方案3】：希望对你有帮助
new_df = df.groupby('datetime')
for hour_group in new_df.groups:
min_ws = new_df.get_group(i)['Sp_mean'].min()
max_ws = new_df.get_group(i)['Sp_mean'].max()
lat = new_df.get_group(i)['x']
lon = new_df.get_group(i)['y']
ws = new_df.get_group(i)['Sp_mean']
plt.figure(figsize=(15,10))
plt.scatter(lon, lat, c=ws,s=300, vmin=min_ws, vmax=max_ws)
plt.title('event' + hour_group.strftime('%Y-%m-%d %H:%M:%S'),
fontweight ='bold',fontsize=18)
plt.show()
【讨论】：
【参考方案4】：如果我理解得很好，您希望将您的数据按 10 分钟分组。如果您的数据集已经过采样，您只需按分钟对数据进行分组并迭代生成的数据帧。
minutes_dfs = df.groupby(df.datetime.map(lambda t: t.minute))
如果还没有采样，你可以分组 10 分钟
minutes_dfs = a.groupby(pd.Grouper(freq='10Min'))
完整代码：
# Example Data Frame
data = 'TimeStamp':['31/10/2017 23:50:00', '31/10/2017 23:50:00', '31/10/2017 23:50:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00'],
'datetime':['31/10/2017 23:50:00', '31/10/2017 23:50:00', '31/10/2017 23:50:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00', '31/10/2017 23:40:00'],
'x':[1, 2, 3, 1, 2, 3, 4, 5, 6, 7, 8, 9],
'y':[9, 8, 7, 9, 8, 7, 6, 5, 4, 3, 2, 1],
'Sp_mean':[6.49, 5.63, 7.07, 7.86, 7.27, 6.59, 6.78, 8.35, 6.3, 5.82, 8.74, 8.94]
df = pd.DataFrame(data)
df['TimeStamp'] = pd.to_datetime(df['TimeStamp'])
df['datetime'] = pd.to_datetime(df['datetime'])
df = df.set_index('datetime')
df['datetime'] = df.index
print(df)
#If data is already sampled
#minutes_dfs = df.groupby(df.datetime.map(lambda t: t.minute))
#Not sampled data
minutes_dfs = a.groupby(pd.Grouper(freq='10Min'))
for min, minutes in minutes_dfs:
t_text = str(min)
#get data into variable
ws = minutes['Sp_mean']
lat = minutes['x']
lon = minutes['y']
col = 0.75
#calc min/max for setting scale on images
min_ws = df['Sp_mean'].min()
max_ws = df['Sp_mean'].max()
plt.figure(figsize=(15,10))
plt.scatter(lon, lat, c=ws,s=300, vmin=min_ws, vmax=max_ws)
plt.title('event' + t_text,fontweight = 'bold',fontsize=18)
plt.show()
【讨论】：
【参考方案5】：简单的解决方案是： 假设 -- df = df.set_index('datetime')... 等
使用：https://numpy.org/doc/stable/reference/arrays.datetime.html
start_date = df.index[0]
min_10 = np.timedelta64(10,'m')
for date in df.index[1::]:
if np.timedelta64(date - start) >= min_10:
start = date
# do your plotting
【讨论】：
以上是关于在日期时间中每 10 分钟绘制一次的主要内容，如果未能解决你的问题，请参考以下文章
将索引转换为日期时间对象后，MatplotLib 无法正确绘制熊猫时间序列 1 分钟数据