def get_should_long(prices: np.array) -> Tuple[float, float]:
# use window to calculate moving avg and moving standard deviation
fast = 9
slow = 20
moving_avg_fast = np.Series(prices).rolling(fast).mean()
moving_avg_slow = np.Series(prices).rolling(slow).mean()
moving_std = np.Series(prices).rolling(slow).std()
bollinger_high = moving_avg_slow + (2 * moving_std)
bollinger_low = moving_avg_slow - (2 * moving_std)
return (moving_avg_fast[-1] - moving_avg_slow[-1]) / prices[-1], (
prices[-1] - bollinger_high[-1])
def beauty_coefficient(c):
"""
This function calculates the sleeping beauty coefficient and awakening time for a publication. See :cite:`ke2015beauty` for details.
Parameters
----------
c : numpy array
The yearly citation counts for the publication.
Returns
----------
B : float
Sleeping Beauty Coefficient
t_a : int
The awakening time
"""
t_m = np.argmax(c)
B_denom = c
B_denom[c == 0] = 1
# :cite:`ke2015beauty` eq 1/2
l_t = ((c[t_m] - c[0]) / t_m * np.arange(c.shape[0]) + c[0] - c) / B_denom
# :cite:`ke2015beauty` eq 2
B = l_t[:(t_m + 1)].sum()
d_denom = np.sqrt((c[t_m] - c[0])**2 + t_m**2)
d_t = np.abs((c[t_m] - c[0]) * np.arange(c.shape[0]) + t_m *
(c[0] - c)) / d_denom
# :cite:`ke2015beauty` eq 3
t_a = np.argmax(d_t[:(t_m + 1)])
return np.Series([B, t_a])
pse.dtype //获取数据类型
pse.astype(np.float64) //修改数据类型，但是pse原本的数据类型不变
计算的特性与数组相同
d={'a':1,'b':2,'c':3,'d':4,'e':5}
# print(pd.Series(d))
'基于索引的数据选择'
print(pd.Series(d,index=list('abcde')))# pandas中数据的选择是基于索引完成的
print(pd.Series(d,index=list('edcba')))# 索引逆序
print(pd.Series(d,index=list('abcabc')))# 索引重复
print(pd.Series(d,index=list('ABCde')))# 索引无法对齐,自动填充缺失值
添加
Series.append()
pse = pd.Series(data=[18, 30, 25, 40], index=list('abcd'), name="user_age_info", dtype=float)
pse.append(np.Series([1,2]),ignore_index=True) //ignore_index为True，则添加后索引在原Series中继续相加，False则为新添加的数据重新从0开始添加索引
除
se = pd.Series(data=[18, 30, 25, 40], index=list('abcd'), name="user_age_info", dtype=float)
pse.drop('a) //不修改源对象
del pse['a'] //删除源对象
inplace=True //参数inplace，表示是否对数据源进行更改
pse.drop_duplicates() //去重 返回带索引值
pse.unique() //去重 以数组形式返回
pse.nunique() //统计非重复数量 与pse.drop_duplicates().count() 等效
判断元素是否存在Serise的值中
18 in pse.values
np.isin() //一个序列是否在另一个序列中
#Filling Missing Data
df
df.fillna(0)
#Calling fillna with a dict you can use a different fill value for each column
df
df.fillna({1: 0.5, 3: -1})
#modify the existing object in place instead of returning a new object
_ = df.fillna(0, inplace=True) #always returns a reference to the filled object
df
#using ffill
df = pd.DataFrame(np.random.randn(6, 3))
df.loc[2:, 1] = NA
df.loc[4:, 2] = NA
df
df.fillna(method='ffill')
df.fillna(method='ffill', limit=2)
#using fillna and passing mean/median
data = np.Series([1., NA, 3.5, NA, 7])
data.fillna(data.mean())
ginf_df = ginf_df[ginf_df['adv_stats'] == True]
ginf_df = ginf_df[['ht', 'at', 'fthg',
'ftag']] # Choose only necessary columns
ginf_df.head(5)
def get_winner(sr):
if sr.fthg > sr.ftag:
return 'home'
elif sr.fthg < sr.ftag:
return 'away'
return 'tie'
# Add winner column
winner_list = []
for index, row in ginf_df.iterrows():
winner_list.append(get_winner(row))
ginf_df = ginf_df.assign(winner=pd.Series(winner_list).values)
# Combine into text dataset
df = events_df.join(ginf_df)
df.head(3)
print(events_df.shape)
print(ginf_df.shape)
print(df.shape)
# Write to file with only necessary columns
df.to_csv(
"drive/Team Drives/Deep Learning Project/ken_cnn/text_dataset_60min.csv",
sep='\t')
import numpy as np
import pandas as pd
s = pd.Series(np.arrange(100,105), ['a','b','c','d','e'])
s.drop('k') #복사본 반환
s
s.drop('k', inplace=True) #바로 적용
s[['a','b']] = [300,900]
s1 = pd.Series(np.Series(np.arrange(100,105))
s1[1:3]
s2['c':'d'] #마지막 포함
corner_kick.append(0)
isFoul.append(0)
isYellowCard1.append(0)
isYellowCard2.append(0)
straight_red_card.append(0)
substitution.append(0)
free_kick_awarded.append(0)
off_sides.append(0)
is_hand_ball.append(0)
penalty_awarded.append(0)
key_pass.append(0)
failed_through_ball.append(0)
sending_off.append(0)
#Create new columns for the 15 additional events.
df["Attempted_Shot"] = pd.Series(attempted_shot)
df["Corner_Kick"] = pd.Series(corner_kick)
df["Foul"] = pd.Series(isFoul)
df["First_Yellow_Card"] = pd.Series(isYellowCard1)
df["Second_Yellow_Card"] = pd.Series(isYellowCard2)
df["Straight_Red_Card"] = pd.Series(straight_red_card)
df["Substitution"] = pd.Series(substitution)
df["Free_Kick_Awarded"] = pd.Series(free_kick_awarded)
df["Off_Sides"] = pd.Series(off_sides)
df["Hand_Ball"] = pd.Series(is_hand_ball)
df["Penalty_Awarded"] = pd.Series(penalty_awarded)
df["Key_Pass"] = pd.Series(key_pass)
df["Failed_Through_Ball"] = pd.Series(failed_through_ball)
df["Sending_Off"] = pd.Series(sending_off)
df["Own_Goal"] = pd.Series(own_goal)
