import pandas as pd
基礎(chǔ)
-
創(chuàng)建
image.png
pd.Series([30, 35, 40], index=['2015 Sales', '2016 Sales', '2017 Sales'], name='Product A')
讀取
- wine_reviews = pd.read_csv("../input/wine-reviews/winemag-data-130k-v2.csv")
有很多參數(shù) 可以有index_col=0
wine_reviews.shape
wine_reviews.head()
- 加載xls文件中的一張表
wic = pd.read_excel("../input/publicassistance/xls_files_all/WICAgencies2013ytd.xls", sheet_name='Total Women')
*SQL
import sqlite3
conn = sqlite3.connect("../input/188-million-us-wildfires/FPA_FOD_20170508.sqlite")
fires = pd.read_sql_query("SELECT * FROM fires", conn)
- 寫(xiě)
- wine_reviews.head().to_csv("wine_reviews.csv")
- wic.to_excel('wic.xlsx', sheet_name='Total Women')
- conn = sqlite3.connect("fires.sqlite")
fires.head(10).to_sql("fires", conn)
- 其他院刁，更多見(jiàn)簡(jiǎn)單手冊(cè) Cheat Sheet
NA空白填充 s.add(s3, fill_value=0)
pd.set_option("display.max_rows", 5)
選擇
- 基于index （row-first, column-second）
reviews.iloc[0] 獲取row 行
reviews.iloc[:, 0] 獲取列column
綜合使用 reviews.iloc[:3, -5:]
- 基于標(biāo)簽 label-based（row-first, column-second）
reviews.loc[:, ['taster_name', 'taster_twitter_handle', 'points']]
loc 0:10,返回11個(gè)，0-10而不是0-9
重設(shè)置index
reviews.set_index("title")
reviews.loc[(reviews.country == 'Italy') & (reviews.points >= 90)]
reviews.loc[reviews.country.isin(['Italy', 'France'])]
reviews.loc[reviews.price.notnull()] # isnull()
range(start, stop[, step])
獲取統(tǒng)計(jì)信息
.describe()
reviews.taster_name.unique()
reviews.points.mean()
reviews.taster_name.value_counts() # 對(duì)于字符串算行，計(jì)算不同種類的出現(xiàn)次數(shù)
map apply
對(duì)于series
review_points_mean = reviews.points.mean()
reviews.points.map(lambda p: p - review_points_mean)
對(duì)于dataframe
def remean_points(srs):
srs.points = srs.points - review_points_mean
return srs
reviews.apply(remean_points, axis='columns')
多減一（默認(rèn)執(zhí)行map/broadcast）
review_points_mean = reviews.points.mean()
reviews.points - review_points_mean
字符串
reviews.country + " - " + reviews.region_1
df.columns = [col.replace(' ', '_').lower() for col in df.columns]
reviews.groupby('points').points.count() 等價(jià)于 reviews['points'].value_counts() （挺有用的）
reviews.groupby('points').price.min()
reviews.groupby(['country', 'province']).apply(lambda df: df.loc[df.points.argmax()]) # 各產(chǎn)地points最大的商品
countries_reviewed.reset_index()
后加內(nèi)容
相關(guān)矩陣
correlation_dataframe.corr()
分箱
