Pandas Visual Analysis

Interactive visual analysis in Jupyter notebooks.


Using pip

To install this package with pip run:

pip install pandas-visual-analysis

Using conda

To install this package with conda run:

conda install -c meffmadd pandas-visual-analysis

From Source

To install this package from source, clone into the repository or download the zip file and run:

python install

JupyterLab Support

Run the following commands to install the necessary JupyterLab extensions:

# renderer support
jupyter labextension install jupyterlab-plotly

# Jupyter widgets extension for
jupyter labextension install @jupyter-widgets/jupyterlab-manager plotlywidget

Note that this requires node to be installed!


Basic Usage

Having a DataFrame, for example:

import pandas as pd
import ssl
ssl._create_default_https_context = ssl._create_unverified_context
url = ""

df = pd.read_csv(url)

you can just pass it to VisualAnalysis to display the default layout:

from pandas_visual_analysis import VisualAnalysis

If you want to explicitly specify which columns of the DataFrame are categorical, just pass the categorical_columns option:

from pandas_visual_analysis import VisualAnalysis
categorical = ["name", "origin", "model_year", "cylinders"]
VisualAnalysis(df, categorical_columns=categorical)

Selection Types

selection types

By default a new selection replaces the old selection, however, it is also possible to add data points to the existing selection by selecting the Additive selection type. By choosing the Subtractive selection, newly selected data points are removed from the selection.

Using DataSource

Instead of passing the DataFrame object directly to VisualAnalysis it is possible to use a DataSource object. This enables linked-brushing across multiple notebook cells if the object is used across cells.

from pandas_visual_analysis import VisualAnalysis, DataSource

data = DataSource(df)

Later you can create a new analysis with the brushing still preserved simply by using the same data object created earlier.


Using Layouts

If you want to specify your own layout, you can do that by passing the layout parameter. The parameter is a list of rows, where each row is in turn a list specifying the Widgets in that row.

from pandas_visual_analysis import VisualAnalysis

    layout=[["Scatter", "Scatter"],

Here, two scatter plots will share the first row while the second row only contains a parallel coordinates plot. In order to see all the possible options you can call the widgets class-method of VisualAnalysis.


This outputs the following list of possible plots:


Any of those can be part of the layout specification. See also: widgets Documentation.

For more advanced features of the VisualAnalysis class see: VisualAnalysis Documentation

Indices and tables