目录
1. DateTime
- date: stores the date (month,day,year)
- time: stores the time (hours,minutes,seconds)
- datetime: stores the date as well as the time (month,day,year,hours,minutes,seconds)
- timedelta: duration between two datetime or date objects
1.1 datetime.date/datetime
- 计算到现在的天数
import datetime
century_start = datetime.date(2000,1,1)
today = datetime.date.today()
today-century_start
- 计算到现在的具体时间
century_start = datetime.datetime(2000,1,1,0,0,0)
time_now = datetime.datetime.now()
time_now - century_start
1.2 datetime.timedelta
datetime.timedelta(days=0, seconds=0, microseconds=0, milliseconds=0, minutes=0, hours=0, weeks=0)
- 可用于datetime和date的加减,不能用于time
today=datetime.date.today()
five_days_later=today+datetime.timedelta(days=5)
now=datetime.datetime.today()
five_minutes_and_five_seconds_later = now + datetime.timedelta(minutes=5,seconds=5)
- 如果要写time的加减,写function
def add_to_time(time_object,time_delta):
import datetime
temp_datetime_object = datetime.datetime(500,1,1,time_object.hour,time_object.minute,time_object.second)
#print(temp_datetime_object)
return (temp_datetime_object+time_delta).time()
1.3 datetime & strings
%y 两位数的年份表示(00-99)
%Y 四位数的年份表示(000-9999)
%m 月份(01-12)
%d 月内中的一天(0-31)
%H 24小时制小时数(0-23)
%I 12小时制小时数(01-12)
%M 分钟数(00=59)
%S 秒(00-59)
%a 本地简化星期名称
%A 本地完整星期名称
%b 本地简化的月份名称
- datetime.strptime(): 从str转化date/datetime/time object
-
date='01-Apr-03' date_object=datetime.datetime.strptime(date,'%d-%b-%y')
但对于delta不成立,所以写function:
bus_travel_time='2:15:30'
hours,minutes,seconds=bus_travel_time.split(':')
x=datetime.timedelta(hours=int(hours),minutes=int(minutes),seconds=int(seconds))
-
datetime.strftime():从datetime转化string
now = datetime.datetime.now()
string_now = datetime.datetime.strftime(now,'%m/%d/%y %H:%M:%S')
2. Data Preprocess
2.1 function
- 对一列的操作
def fix_zip(input_zip):
try:
input_zip = int(float(input_zip))
except:
try:
input_zip = int(input_zip.split('-')[0])
except:
return np.NaN
if input_zip < 10000 or input_zip > 12000:
return np.NaN
return str(input_zip)
df['Incident Zip'] = df['Incident Zip'].apply(fix_zip)
- 对一整个df的操作
def GroupColFunc(df, ind, col):
if df[col].loc[ind] < 0:
return 'Negative'
else:
return 'Positive'
grouped = people.groupby(lambda x: GroupColFunc(people, x, 'a'))
- 对其中某几列操作
def foo(x,y):
return x+y
df_sample['new_col'] = df_sample.apply(lambda x: foo(x['A'],x['B']),axis=1)
#axis = 1 means extract data from rows of the dataframe. by default, it extracts from columns. (Set axis to 0 and see what happens)
2.2 Null值
- 去掉zip=Nan的行
df = df[df['Incident Zip'].notnull()]
df = df[(df['Latitude'].notnull()) & (df['Longitude'].notnull()) & (df['Closed Date'].notnull())]
- 去除含有空值的行或列
-
df.dropna(axis=0, how='any') #axis=0行;axis=1列; #how=any有元素缺失就删除;how=all全部缺失才删除 df.fillna(value=None,method=None,axis=None) #value和method二选一:value意为用什么值去填充; #method:ffill前一个值;backfill/bfill后一个值; #axis=1横向的前后值替换;axis=0上下的值替换 df.isna() df.isnull() #判断是否为空值
2.3 两个df合并与重塑
-
concat: 0为上下合并;1为左右合并
pd.concat([df1,df2], axis=0, join='outer',keys=['x','y'])
objs: series,dataframe或者是panel构成的序列lsit
axis: 需要合并链接的轴,0是行,1是列
join:连接的方式 inner,或者outer
要在相接的时候在加上一个层次的key来识别数据源自于哪张表,可以增加key参数;
join='inner'交集;join='outer'并集
- merge:通过key左右合并列
result = pd.merge(left, right, on=['key1', 'key2'])
how='left'保留左表的所有数据;how=‘outer’保留两表所有信息;how=‘inner’保留公共部分
- join:左右合并
df1.join(df2) #左右合并 没有的就填Nan
df1.join(df2,how='right') #取共同行
2.4 其他
- where判断
df['UP']=np.where(df['Close']>df['Open'],1,0) #增加一列,判断是否up
- percentage change
df['Close'].pct_change() #One timeperiod percent change
- statistical用法
grouped.mean()
grouped.std()
grouped.count()
grouped.cumcount()
- 定位
df.loc[date(2018,9,21)]
df.iloc[0]
- 命名
df = pd.DataFrame({"aa":[1,2,3], "bb": [4,5,6]})
df.columns=["a", "b"] # rename all columns
df.rename(columns={"a":"a1"}) #选择一部分重命名