Pandas是數(shù)據(jù)挖掘常見(jiàn)的工具，掌握使用過(guò)程中的函數(shù)是非常重要的。本文將借助可視化的過(guò)程，講解Pandas的各種操作。

sort_values

(dogs[dogs['size']=='medium']
.sort_values('type')
.groupby('type').median()
)

執(zhí)行步驟：

• size列篩選出部分行
• 然后將行的類(lèi)型進(jìn)行轉(zhuǎn)換
• 按照type列進(jìn)行分組，計(jì)算中位數(shù)

selecting a column

dogs['longevity']

groupby + mean

dogs.groupby('size').mean()

• 將數(shù)據(jù)按照size進(jìn)行分組
• 在分組內(nèi)進(jìn)行聚合操作

grouping multiple columns

dogs.groupby(['type','size'])

groupby + multi aggregation

(dogs
.sort_values('size')
.groupby('size')['height']
.agg(['sum','mean','std'])
)

執(zhí)行步驟

• 按照size列對(duì)數(shù)據(jù)進(jìn)行排序
• 按照size進(jìn)行分組
• 對(duì)分組內(nèi)的height進(jìn)行計(jì)算

filtering for columns

df.loc[:,df.loc['two']<=?20]

filtering for rows

dogs.loc[(dogs['size']=='medium')&(dogs['longevity']>12),'breed']

dropping columns

dogs.drop(columns=['type'])

joining

ppl.join(dogs)

merging

ppl.merge(dogs,left_on='likes',right_on='breed',how='left')

pivot table

dogs.pivot_table(index='size',columns='kids',values='price')

melting

dogs.melt()

pivoting

dogs.pivot(index='size',columns='kids')

stacking column index

dogs.stack()

unstacking row index

dogs.unstack()

resetting index

dogs.reset_index()

setting index

dogs.set_index('breed')