数据整理(Data Wrangling)
数据整理(Data Wrangling)可归纳为以下三步:
- 数据收集(Gather)
- 数据评估(Assess)
- 数据清理(Clean)
数据收集(Gather)
收集数据的方式有很多,最简单、最常见的是下载现成的数据。比如从kaggle上下载数据集。
但为了可扩展性(Scalability)和重复性(Reproducibility),有时需要以编程的(Programmatically)方式下载。比如需要下载的文件量较大,有成百上千个,甚至位于不同页面。
从网上爬取数据。比如爬知乎,爬豆瓣。
从各种API获得数据。比如电影数据API,股票数据API,Twitter数据API,等等。
数据评估(Assess)
可以从两方面进行:质量(Quality),整洁度(Tidiness)
质量(Quality)
低质量数据常被称为脏数据(dirty data),比如:
- 数据丢失,缺值。
- 数据无效。
- 数据不准确。
- 数据不一致,比如使用不同的长度单位(英寸和厘米)。
整洁度(Tidiness)
不整洁数据常被称为杂乱数据(messy data),是统计学家、教授和全能数据专家 Hadley Wickham 提出的概念。
A dataset is messy or tidy depending on how rows, columns, and tables are matched up with observations, variables, and types. In tidy data:
- Each variable forms a column.
- Each observation forms a row.
- Each type of observational unit forms a table.
数据清理(Clean)
分为手工清理和程序清理。
程序清理:
- Define: convert our assessments into defined cleaning tasks. These definitions also serve as an instruction list so others (or yourself in the future) can look at your work and reproduce it.
- Code: convert those definitions to code and run that code.
- Test: test your dataset, visually or with code, to make sure your cleaning operations worked.
Always make copies of the original pieces of data before cleaning!
Reassess and Iterate
- After cleaning, always reassess and iterate on any of the data wrangling steps if necessary.
Store (Optional)
- Store data, in a file or database for example, if you need to use it in the future.