Recommendation of Altair! Data visualization with Python

About this article

Introducing ** Altair **, Python's data visualization library. Since there is little information in Japanese, I will write an article for missionary purposes. If you are comfortable with English, it is easiest to see the Official Page.

Installation

You can easily install it with the pip command. vega_datasets will be used later, so let's install it together.

pip install altair vega_datasets

I used Google Colaboratory and it worked from the beginning without installation.

Advance preparation

Load the library and dataset below.

import altair as alt
from vega_datasets import data
iris = data.iris()

Altair is good at working with Pandas, and ʻiris` is a Pandas DataFrame.

Basic visualization

The following is the assumed code to be visualized directly with Jupyter etc. If you want to output in html, add .save ("filename.html ") to the end. You don't need to have .interactive (), but if you write it, you will be able to move the graph. ** This article is a normal image, so if you want to move it around, please use here **.

Example ① Scatter plot

Specify the x-axis and y-axis values as follows: You can also write ʻalt.X () `like a comment, and use this for complicated visualization.

alt.Chart(iris).mark_point().encode(
    x="sepalLength", # alt.X("sepalLength"),
    y="sepalWidth", # alt.Y("sepalWidth"),
    color="species"
).interactive()

Example ② Bar graph

The point is to take the average for each species with ʻaverage ()`. You can perform various operations other than average, and you can check the list at here.

alt.Chart(iris).mark_bar().encode(
    x="average(sepalLength)", # alt.X("sepalLength", aggregate="average"),
    y="species", # alt.Y("species"),
).interactive()

The point is like this. If you replace the make_xxxxx part with make_line, you can easily draw a line graph. If you have a problem, you can usually solve it by searching for a similar graph from Gallery on the official page.

TIPS (may be added)

Information display by mouse over

The information specified by the tooltip argument is displayed by mouse over.

alt.Chart(iris).mark_point().encode(
    x="sepalLength",
    y="sepalWidth",
    color="species",
    tooltip=["sepalLength", "sepalWidth", "petalLength", "petalWidth", "species"]
).interactive()

I don't want to write the axis from 0

Quantitative data is basically visualized including 0. The graph above clearly indicates that it does not contain 0 with zero = False.

alt.Chart(iris).mark_point().encode(
    alt.X("sepalLength", scale=alt.Scale(zero=False)),
    alt.Y("sepalWidth", scale=alt.Scale(zero=False)),
    color="species"
).interactive()

Data type specification

Nominal scales are often integers, aren't they? In that case, specify that it is a nominal scale, such as species_int: N. By the way, the ordinal scale is : O, and for quantitative data it is: Q. Details can be found in the official documentation here.

#Convert to integer value(setosa: 0, versicolor: 1, virginica: 2)
iris["species_int"] = [["setosa", "versicolor", "virginica"].index(x) for x in iris["species"]]

#Correct example
alt.Chart(iris).mark_point().encode(
    x="sepalLength",
    y="sepalWidth",
    color="species_int:N"
).interactive()

By the way, without : N, it will be as follows.

Supports MaxRowError

I get angry when I exceed 5000 lines. Execute the following referring to the information in here.

alt.data_transformers.disable_max_rows()

Finally

It can be described simply and is convenient for exploratory analysis. If there is a drawback, html output is easy, but png output seems to be a little difficult. FYI!