[PYTHON] Pandas reverse lookup memo

I will put together things that I will use again later (Note: Please note that there is nothing new about the content of the personal memo article)

Introduction

Data set preparation

import pandas as pd
from sklearn import datasets
iris = datasets.load_iris()
columns = list(map(lambda x: ' '.join(x.split(' ')[:2]), iris.feature_names))
df = pd.DataFrame(iris.data, columns=columns)
df['target'] = iris.target_names[iris.target]
df.head()

View the number of data per label

Overview

df.groupby('target').size().to_frame.plot.barh()

Execution result:

Commentary

• I'm using to_frame to use pandas.DataFrame.plot.barh
• COUNT GROUP BY by size () for each value in the specified column
• Both bar and barh are possible, but I personally like barh, which has easy-to-read labels.
• If not graphed, the DataFrame looks like this:

See a color-coded scatter plot for each label

Overview

for key, indices in df.groupby('target').groups.items():
    x = df.loc[indices]['sepal length']
    y = df.loc[indices]['petal length']
    plt.scatter(x, y, label=key, alpha=0.4)
plt.legend()
plt.show()

Execution result:

Commentary

• df.groupby ('target'). groups gives a dictionary of the form{specified column values: index}containing the specified column values.
• It can be treated as an iterator that returns keys and values in order with .items () attached to the dictionary type in Python standard.
• Others are displayed using matplotlib.