Pandas
Pandas is a Python library used to analyze data and work with data sets. The name is derived from both "Panel Data" and "Python Data Analysis". [1]
Pandas is used to cleanse bad data sets, find correlations, and analyze the dataset using both Data Frames and Series.
To install the Pandas library, install onto your version of Python from command line using the following command:
pip install pandas
To import the Pandas library into your scripts, use the following code to have access to it as a reference:
In Pandas, the libraries has a property called Series. They are one dimensional arrays which act as a column from a database, and any data type can be stored inside. A singular "cell" of data can be accessed using either the numerical index, or by referencing them using any created labels associated with that index.
Another concept in Pandas are called DataFrames. These act similarly to tables and are closely related to Series, much like a column-table relationship in relational databases and are multidimensional arrays. Big data sets can be easily stored inside a Series structure, and the module supports direct conversion from common file types such as both JSON and CSV extensions. The info() method can be used to print info about the dataset, such as total number of columns, rows, non-null values in each column, and datatype of each column, which is extremely useful for knowing if the dataset needs more cleanup.
[1] Source: W3Schools - Pandas Introduction [1] Source: W3Schools - Pandas IntroductionThis module works in conjunction with the Pandas library, and creates a detailed profile report of your dataset. The following information is generated when using the df.profile_report()function:
- checks the datatypes of the columns
- Any unique, missing, or frequently occurring values
- Any duplicate rows of data
- Quantile statistics
- mean, median, mode, standard deviation, coefficient of variance
- creates histograms
- Underscores correlations between variables within the dataset
- A textual review of the dataset [1]
Here is a sample dataset and a corresponding generated report from that dataset:
data setreport
pip install pandas-profiling
To use the pandas-profiling module, use the following command in your scripts:
Here is an example of how to generate a report:
[1] Source: Pandas-Profiling: Introduction
Creating a DataFrame
Output:
Missing data
If you encounter datasets with missing data, you can use the numpy module to fill in any cells with NaN as a value to what you choose, though most commonly, zero is recommended to avoid contaminating the data.
Output:
You can also use the interpolate() function to interpolate these values.
The other option as well is to drop any NaN values
Output:
Iteration
You can also loop through a dataset's rows or columns as well. Iterating though a data frame's columns, is fairly straight forward. You can treat the DataFrame as a list, and select a particular index to select a certain column. To iterate and select rows, treat the dataframe like a 2D array and use .iterrows() to loop through.
Columns
Output:
Rows
Output:
Source: Towards Data Science: A Simple Guide to Pandas DataFrames
Description
Covers why Python is so widely used in Data Science, as well as commonly used
Python
libraries used in Python in data science such as:
- Numpy
- Pandas
- Matplotlib
- Sklearn
Description
Video tutorial on using Pandas for Data Science, such as reading the file and explains how to read one in based on conditioning, as well as number of common operations used by the library.
Description
A video series on using the Pandas library including interactive Python programming quizzes.