data-view 0.2.0

Automated view of dataset

Description

Requirements

• python >= 3.6
• numpy >= 1.16
• pandas >= 0.23
• matplotlib >= 3.02
• IPython >= 7.2.0

---

Module, provides the function view, which displays general information on the data:

• Result of method info()
• Result of method describe()
• for numeric / categorical variables - The number of nulls in the data (amount and percentage for each column)
• grid of histograms for numeric variables
• Top-5 of the most frequent categorical variables (for each)
• Split count statistic by category with less 5 unique values
• matrix correlation
• dataframe of higher correlation pair (over 0.7)

Parameters (function view):

• d - table with data
• only_numeric - True / False, default: True. True - information output only by numerical variables, False - information output by numerical and categorical variables.
• full_stats - True / False, default: False. False - output information on numerical variables without interquartile range, data boundaries without outliers, True - complete output with IQR, min and max by IQR and amount of outliers.
• histograms - True / False, default: True. True - output with building histograms for numerical variables, False - without building histograms

Top-5 elements of categorical variables

The postfix (_name / _count) is assigned to the name of the data column to create top-5 elements table:

• _name - category name
• _count - number of unique elements in this category. If there are less than 5 unique elements, then the values ​​in the _count field are filled -1

Count statistic of numbers variables split by some categorical variables

• if number of unique elements in category variable less 5, then show count statistic

Correlation matrix

• Correlation matrix with heatmap (pearson corr index)

Install

$ pip install data_view

Usage

$ python3

from data_view import *

d = pd.DataFrame(np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]]), columns=['a', 'b', 'c'])

view(d, only_numeric=True, histograms=False)