数据清洗简介

在 PostgreSQL 数据库中清理数据

Darryl Reeves, Ph.D.

Industry Assistant Professor, New York University

为何数据清洗重要？

数据很杂乱
在分析前通常需要清洗
善用列类型约束
课程聚焦于无法/不宜采用防御式方法的情形

清洗字符串数据

字符串数据：多、灵活，且常常混乱

 name                            | grade |            inspection_type              | census_tract | ... 
 --------------------------------+-------+-----------------------------------------+--------------+-----
 ...                             | ...   | ...                                     | ...          | ...
 EMPANADAS MONUMENTAL            | B     | Cycle Inspection  /  Re-inspection      | 26900        | ...
 ALPHONSO'S PIZZERIA & TRATTORIA | A     | Cycle Inspection  /  Initial Inspection | 202          | ...
 THE SPARROW TAVERN              | A     | Cycle Inspection  /  Initial Inspection | 12500        | ...
 BURGER KING                     | A     | Cycle Inspection  /  Re-inspection      | 86400        | ...
 ASTORIA PIZZA                   | B     | Cycle Inspection  /  Re-inspection      | 6300         | ...
 ...                             | ...   | ...                                     | ...          | ...

清洗字符串数据

 name                            | grade |            inspection_type              | census_tract | ... 
 --------------------------------+-------+-----------------------------------------+--------------+-----
 ...                             | ...   | ...                                     | ...          | ...
 EMPANADAS MONUMENTAL            | B     | Cycle Inspection  /  Re-inspection      | 26900        | ...
 ALPHONSO'S PIZZERIA & TRATTORIA | A     | Cycle Inspection  /  Initial Inspection | 202          | ...
 THE SPARROW TAVERN              | A     | Cycle Inspection  /  Initial Inspection | 12500        | ...
 BURGER KING                     | A     | Cycle Inspection  /  Re-inspection      | 86400        | ...
 ASTORIA PIZZA                   | B     | Cycle Inspection  /  Re-inspection      | 6300         | ...
 ...                             | ...   | ...                                     | ...          | ...

规范 name 的大小写
去除 inspection_type 中多余分隔空格
使 census_tract 长度一致

清洗字符串数据

 name                            | grade |            inspection_type              | census_tract | ... 
 --------------------------------+-------+-----------------------------------------+--------------+-----
 ...                             | ...   | ...                                     | ...          | ...
 EMPANADAS MONUMENTAL            | B     | Cycle Inspection  /  Re-inspection      | 26900        | ...
 ALPHONSO'S PIZZERIA & TRATTORIA | A     | Cycle Inspection  /  Initial Inspection | 202          | ...
 THE SPARROW TAVERN              | A     | Cycle Inspection  /  Initial Inspection | 12500        | ...
 BURGER KING                     | A     | Cycle Inspection  /  Re-inspection      | 86400        | ...
 ASTORIA PIZZA                   | B     | Cycle Inspection  /  Re-inspection      | 6300         | ...
 ...                             | ...   | ...                                     | ...          | ...

 name                            | grade |           inspection_type             | census_tract | ... 
 --------------------------------+-------+---------------------------------------+--------------+-----
 ...                             | ...   | ...                                   | ...          | ...
 Empanadas Monumental            | B     | Cycle Inspection / Re-inspection      | 026900       | ...
 Alphonso'S Pizzeria & Trattoria | A     | Cycle Inspection / Initial Inspection | 000202       | ...
 The Sparrow Tavern              | A     | Cycle Inspection / Initial Inspection | 012500       | ...
 Burger King                     | A     | Cycle Inspection / Re-inspection      | 086400       | ...
 Astoria Pizza                   | B     | Cycle Inspection / Re-inspection      | 006300       | ...
 ...                             | ...   | ...                                   | ...          | ...

使用 INITCAP() 函数

INITCAP(input_string) - 修正大小写

SELECT INITCAP('HELLO FRIEND!');

Hello Friend!

使用 REPLACE() 函数

REPLACE(input_string, to_replace, replacement) - 将一个文本值替换为另一个

SELECT REPLACE('180 Main Street', 'Street', 'St');

180 Main St

使用 LPAD() 函数

LPAD(input_string, length [, fill_value]) - 在字符串前填充文本

SELECT LPAD('123', 7, 'X');

XXXX123

构建字符串清洗查询

SELECT
  INITCAP(name) as name,
  grade,
  REPLACE(inspection_type, '  /  ', ' / ') as inspection_type, 
  LPAD(census_tract, 6, '0') as census_tract
FROM
  restaurant_inspection;

 name                            | grade |           inspection_type             | census_tract | ... 
 --------------------------------+-------+---------------------------------------+--------------+-----
 ...                             | ...   | ...                                   | ...          | ...
 Empanadas Monumental            | B     | Cycle Inspection / Re-inspection      | 026900       | ...
 Alphonso'S Pizzeria & Trattoria | A     | Cycle Inspection / Initial Inspection | 000202       | ...
 The Sparrow Tavern              | A     | Cycle Inspection / Initial Inspection | 012500       | ...
 Burger King                     | A     | Cycle Inspection / Re-inspection      | 086400       | ...
 Astoria Pizza                   | B     | Cycle Inspection / Re-inspection      | 006300       | ...
 ...                             | ...   | ...                                   | ...          | ...

让我们来练习！

在 PostgreSQL 数据库中清理数据