|NO.Z.00050|——————————|BigDataEnd|——|Hadoop&Python.v14|——|Arithmetic.v14|Pandas数据分析库：Pandas时间序列|
一、时间序列
Walter Savage Landor:strove with none,for none was worth my strife.Nature I loved and, next to Nature, Art:I warm'd both hands before the fire of life.It sinks, and I am ready to depart ——W.S.Landor
### --- 时间戳操作
~~~ # 创建?法
pd.Timestamp('2020-8-24 12') # 时刻数据
pd.Period('2020-8-24',freq = 'M') # 时期数据
index = pd.date_range('2020.08.24',periods=5,freq = 'M') # 批量时刻数据
pd.period_range('2020.08.24',periods=5,freq='M') # 批量时期数据
ts = pd.Series(np.random.randint(0,10,size = 5),index = index) # 时间戳索引Series
~~~ # 转换?法
pd.to_datetime(['2020.08.24','2020-08-24','24/08/2020','2020/8/24'])
pd.to_datetime([1598582232],unit='s')
dt = pd.to_datetime([1598582420401],unit = 'ms') # 世界标准时间
dt + pd.DateOffset(hours = 8) # 东?区时间
dt + pd.DateOffset(days = 100) # 100天后?期### --- 时间戳索引
index = pd.date_range("2020-8-24", periods=200, freq="D")
ts = pd.Series(range(len(index)), index=index)
~~~ # str类型索引
ts['2020-08-30'] # ?期访问数据
ts['2020-08-24':'2020-09-3'] # ?期切?
ts['2020-08'] # 传?年?
ts['2020'] # 传?年
~~~ # 时间戳索引
ts[pd.Timestamp('2020-08-30')]
ts[pd.Timestamp('2020-08-24'):pd.Timestamp('2020-08-30')] # 切?
ts[pd.date_range('2020-08-24',periods=10,freq='D')]
~~~ # 时间戳索引属性
ts.index.year # 获取年
ts.index.dayofweek # 获取星期?
ts.index.weekofyear # ?年中第?个星期?### --- 时间序列常??法
### --- 在做时间序列相关的?作时，经常要对时间做?些移动/滞后、频率转换、采样等相关操作，我们来看下这些操作如何使?
index = pd.date_range('8/1/2020', periods=365, freq='D')
ts = pd.Series(np.random.randint(0, 500, len(index)), index=index)
~~~ # 1、移动
ts.shift(periods = 2) # 数据后移
ts.shift(periods = -2) # 数据前移
~~~ # ?期移动
ts.shift(periods = 2,freq = pd.tseries.offsets.Day()) # 天移动
ts.tshift(periods = 1,freq = pd.tseries.offsets.MonthOffset()) # ?移动
~~~ # 2、频率转换
ts.asfreq(pd.tseries.offsets.Week()) # 天变周
ts.asfreq(pd.tseries.offsets.MonthEnd()) # 天变?
ts.asfreq(pd.tseries.offsets.Hour(),fill_value = 0) # 天变?时，?少变多，fill_value为填充值
~~~ # 3、重采样
~~~ # resample 表示根据?期维度进?数据聚合，可以按照分钟、?时、?作?、周、?、年等来作为?期维度
ts.resample('2W').sum() # 以2周为单位进?汇总
ts.resample('3M').sum().cumsum() # 以季度为单位进?汇总
~~~ # 4、DataFrame重采样
d = dict({'price': [10, 11, 9, 13, 14, 18, 17, 19],
'volume': [50, 60, 40, 100, 50, 100, 40, 50],
'week_starting':pd.date_range('24/08/2020',periods=8,freq='W')})
df1 = pd.DataFrame(d)
df1.resample('M',on = 'week_starting').apply(np.sum)
df1.resample('M',on = 'week_starting').agg({'price':np.mean,'volume':np.sum})
days = pd.date_range('1/8/2020', periods=4, freq='D')
data2 = dict({'price': [10, 11, 9, 13, 14, 18, 17, 19],
'volume': [50, 60, 40, 100, 50, 100, 40, 50]})
df2 = pd.DataFrame(data2,
index=pd.MultiIndex.from_product([days,['morning','afternoon']]))
df2.resample('D', level=0).sum()### --- 时区表示
index = pd.date_range('8/1/2012 00:00', periods=5, freq='D')
ts = pd.Series(np.random.randn(len(index)), index)
import pytz
pytz.common_timezones # 常?时区
~~~ # 时区表示
ts = ts.tz_localize(tz='UTC')
~~~ # 转换成其它时区
ts.tz_convert(tz = 'Asia/Shanghai')Walter Savage Landor:strove with none,for none was worth my strife.Nature I loved and, next to Nature, Art:I warm'd both hands before the fire of life.It sinks, and I am ready to depart ——W.S.Landor