import pandas as pd
import altair as alt
import os
alt.renderers.enable('notebook')
alt.renderers.enable('mimetype')
[docs]def plot_distributions(data, bins = 40, dist_cols=None, class_label=None):
"""This function creates numerical distribution plots on either all the numeric columns or the ones provided to it
Parameters
----------
data : pandas.DataFrame
The dataframe for which distribution plot has to be created
bins : int
The number of bins for histogram plot
dist_cols : list, optional
The subset of numeric columns for which the histogram plots have to be generated
class_label : str, optional
The name of the target column only in case of classification dataset. For regression dataset, it is not required
Returns
-------
chart_numeric
The Altair object for the plot
Examples
-------
>>> plot_distributions(data)
>>> plot_distributions(data, dist_cols=['loudness', 'acousticness'], class_label='target')
"""
if data is None:
raise ValueError("Required arg 'data' cannot be empty")
if not isinstance(data, pd.DataFrame):
raise TypeError("Please enter data of type pd.DataFrame")
if not isinstance(bins, int):
raise TypeError("bins should be of type int")
if dist_cols is not None:
if not isinstance(dist_cols, list):
raise TypeError("The entered dist_cols should be of 'list' type")
else:
numeric_features = dist_cols
else:
numeric_features = list(data.select_dtypes('number').columns)
if class_label is None:
chart_numeric = alt.Chart(data).mark_bar().encode(
alt.X(alt.repeat(), type='quantitative', bin=alt.Bin(maxbins=bins)),
y=alt.Y('count()')
).properties(
width=300,
height=200
).repeat(
numeric_features, columns = 4
)
else:
if not type(class_label) == str:
raise TypeError("`class_label` should be of string type")
else:
chart_numeric = alt.Chart(data).mark_bar(opacity=0.6).encode(
alt.X(alt.repeat(), type='quantitative', bin=alt.Bin(maxbins=bins)),
y=alt.Y('count()', stack=False),
color=class_label+':N'
).properties(
width=300,
height=200
).repeat(
numeric_features, columns = 4
)
return chart_numeric
[docs]def plot_corr(data, corr='spearman'):
"""This function creates correlation plot for all the columns in the dataframe
Parameters
----------
data : pandas.DataFrame
The dataframe for which distribution plot has to be created
corr : str
The correlation method, which can be among 'spearman', 'kendall' or 'pearson'
The default value is spearman
Returns
-------
corr_plot
The Altair object for the plot
Examples
-------
>>> plot_corr(data)
>>> plot_corr(data, corr='kendall')
"""
if data is None :
raise ValueError("Required arg 'data' cannot be empty")
if not isinstance(data, pd.DataFrame):
raise TypeError("Please enter data of type pd.DataFrame")
if corr == '':
raise ValueError("'corr' cannot be empty")
if corr not in ['spearman', 'pearson', 'kendall']:
raise ValueError("corr should be one of these: 'spearman', 'pearson', 'kendall'")
corr_df= data.corr(corr).stack().reset_index(name='corr')
corr_plot = alt.Chart(corr_df, title='Correlation Plot among all features and target').mark_rect().encode(
x=alt.X('level_0', title='All features'),
y=alt.Y('level_1', title='All features'),
tooltip='corr',
color=alt.Color('corr', scale=alt.Scale(domain=(-1, 1), scheme='purpleorange')))
return corr_plot
[docs]def plot_splom(data, pair_cols=None):
"""This function creates SPLOM plot for all the numeric columns in the dataframe or the ones passed by the user
Parameters
----------
data : pandas.DataFrame
The dataframe for which distribution plot has to be created
pair_cols : list
The list of dataframe columns, for which correlation plot is to be generated
Returns
-------
splom_chart
The Altair object for the plot
Examples
-------
>>> plot_splom(data)
>>> plot_splom(data, pair_cols=['loudness', 'acousticness', 'energy'])
"""
if data is None :
raise ValueError("Required arg 'data' cannot be empty")
if not isinstance(data, pd.DataFrame):
raise TypeError("Please enter data of type pd.DataFrame")
if pair_cols==[]:
raise ValueError("pair_cols should not be empty list")
if pair_cols is None:
pair_cols = list(data.select_dtypes('number').columns)
else:
if not isinstance(pair_cols, list):
raise TypeError("The entered pair_cols should be of 'list' type")
splom_chart = alt.Chart(data).mark_point(opacity=0.3, size=10).encode(
alt.X(alt.repeat('row'), type='quantitative'),
alt.Y(alt.repeat('column'), type='quantitative')
).properties(
width=200,
height=200
).repeat(
column=pair_cols,
row=pair_cols
)
return splom_chart