Python可视化：Seaborn库热力图使用进阶

阅读量：4616 次

发布时间：2019-06-09

本文共 8228 字，大约阅读时间需要 27 分钟。

前言

在日常工作中，经常可以见到各种各种精美的热力图，热力图的应用非常广泛，下面一起来学习下Python的Seaborn库中热力图（heatmap）如何来进行使用。

本次运行的环境为：

windows 64位系统

python 3.5

jupyter notebook

1 构造数据

import seaborn as snsimport pandas as pdimport numpy as npimport matplotlib.pyplot as plt% matplotlib inline

region = ['Albania', 'Algeria', 'Angola', 'Argentina', 'Armenia', 'Azerbaijan',       'Bahamas', 'Bangladesh', 'Belize', 'Bhutan', 'Bolivia',       'Bosnia and Herzegovina', 'Brazil', 'Burkina Faso', 'Burundi',       'Cambodia', 'Cameroon', 'Cape Verde', 'Chile', 'China', 'Colombia',       'Costa Rica', 'Cote d Ivoire', 'Cuba', 'Cyprus',       "Democratic People's Republic of Korea",       'Democratic Republic of the Congo', 'Dominican Republic', 'Ecuador',       'Egypt', 'El Salvador', 'Equatorial Guinea', 'Ethiopia', 'Fiji',       'Gambia', 'Georgia', 'Ghana', 'Guatemala', 'Guyana', 'Honduras']kind = ['Afforestation & reforestation', 'Biofuels', 'Biogas',        'Biomass', 'Cement', 'Energy efficiency', 'Fuel switch',       'HFC reduction/avoidance', 'Hydro power',        'Leak reduction', 'Material use', 'Methane avoidance',                    'N2O decomposition', 'Other renewable energies',       'PFC reduction and substitution','PV',       'SF6 replacement', 'Transportation', 'Waste gas/heat utilization',      'Wind power']

print(len(region))print(len(kind))

np.random.seed(100)arr_region = np.random.choice(region, size=(10000,))list_region = list(arr_region)arr_kind = np.random.choice(kind, size=(10000,))list_kind = list(arr_kind)values = np.random.randint(50, 1000, 10000)list_values = list(values)df = pd.DataFrame({
    'region':list_region,                  'kind': list_kind,                  'values':list_values})df.head()

pt = df.pivot_table(index='kind', columns='region', values='values', aggfunc=np.sum)pt.head()

f, ax = plt.subplots(figsize = (10, 4))cmap = sns.cubehelix_palette(start = 1, rot = 3, gamma=0.8, as_cmap = True)sns.heatmap(pt, cmap = cmap, linewidths = 0.05, ax = ax)ax.set_title('Amounts per kind and region')ax.set_xlabel('region')ax.set_ylabel('kind')f.savefig('sns_heatmap_normal.jpg', bbox_inches='tight')# ax.set_xticklabels(ax.get_xticklabels(), rotation=-90)

2 Seaborn的heatmap各个参数介绍

seaborn.heatmap

seaborn.heatmap(data, vmin=None, vmax=None, cmap=None, center=None, robust=False, annot=None, fmt=’.2g’, annot_kws=None, linewidths=0, linecolor=’white’, cbar=True, cbar_kws=None, cbar_ax=None, square=False, ax=None, xticklabels=True, yticklabels=True, mask=None, **kwargs)

data：矩阵数据集，可以使numpy的数组（array），如果是pandas的dataframe，则df的index/column信息会分别对应到heatmap的columns和rows

linewidths,热力图矩阵之间的间隔大小

vmax,vmin, 图例中最大值和最小值的显示值，没有该参数时默认不显示

2.1 cmap

cmap：matplotlib的colormap名称或颜色对象；如果没有提供，默认为cubehelix map (数据集为连续数据集时) 或 RdBu_r (数据集为离散数据集时)

f, (ax1,ax2) = plt.subplots(figsize = (10, 8),nrows=2)# cubehelix map颜色cmap = sns.cubehelix_palette(start = 1.5, rot = 3, gamma=0.8, as_cmap = True)sns.heatmap(pt, linewidths = 0.05, ax = ax1, vmax=15000, vmin=0, cmap=cmap)ax1.set_title('cubehelix map')ax1.set_xlabel('')ax1.set_xticklabels([]) #设置x轴图例为空值ax1.set_ylabel('kind')# matplotlib colormapsns.heatmap(pt, linewidths = 0.05, ax = ax2, vmax=15000, vmin=0, cmap='rainbow') # rainbow为 matplotlib 的colormap名称ax2.set_title('matplotlib colormap')ax2.set_xlabel('region')ax2.set_ylabel('kind')f.savefig('sns_heatmap_cmap.jpg', bbox_inches='tight')

2.2 center

center:将数据设置为图例中的均值数据，即图例中心的数据值；通过设置center值，可以调整生成的图像颜色的整体深浅；设置center数据时，如果有数据溢出，则手动设置的vmax、vmin会自动改变

f, (ax1,ax2) = plt.subplots(figsize = (10, 8),nrows=2)cmap = sns.cubehelix_palette(start = 1.5, rot = 3, gamma=0.8, as_cmap = True)sns.heatmap(pt, linewidths = 0.05, ax = ax1, vmax=15000, vmin=0, cmap=cmap, center=None )# center为None时，由于最小值为0，最大值为15000，相当于center值为vamx和vmin的均值，即7500ax1.set_title('center=None')ax1.set_xlabel('')ax1.set_xticklabels([]) #设置x轴图例为空值ax1.set_ylabel('kind')sns.heatmap(pt, linewidths = 0.05, ax = ax2, vmax=15000, vmin=0, cmap=cmap, center=3000 ) # 由于均值为2000，当center设置为3000时，大部分数据会比7500大，所以center=3000时，生成的图片颜色要深# 设置center数据时，如果有数据溢出，则手动设置的vmax或vmin会自动改变ax2.set_title('center=3000')ax2.set_xlabel('region')ax2.set_ylabel('kind')f.savefig('sns_heatmap_center.jpg', bbox_inches='tight')

2.3 robust

f, (ax1,ax2) = plt.subplots(figsize = (10, 8),nrows=2)cmap = sns.cubehelix_palette(start = 1.5, rot = 3, gamma=0.8, as_cmap = True)sns.heatmap(pt, linewidths = 0.05, ax = ax1, cmap=cmap, center=None, robust=False )# robust默认为Falseax1.set_title('robust=False')ax1.set_xlabel('')ax1.set_xticklabels([]) #设置x轴图例为空值ax1.set_ylabel('kind')sns.heatmap(pt, linewidths = 0.05, ax = ax2, cmap=cmap, center=None, robust=True ) # If True and vmin or vmax are absent, the colormap range is computed with robust quantiles instead of the extreme values.ax2.set_title('robust=True')ax2.set_xlabel('region')ax2.set_ylabel('kind')f.savefig('sns_heatmap_robust.jpg', bbox_inches='tight')

2.4 mask

f, (ax1,ax2) = plt.subplots(figsize = (10, 8),nrows=2)cmap = sns.cubehelix_palette(start = 1.5, rot = 3, gamma=0.8, as_cmap = True)p1 = sns.heatmap(pt, linewidths = 0.05,ax=ax1, vmax=15000, vmin=0, cmap=cmap, center=None, robust=False, mask=None )# robust默认为Falseax1.set_title('mask=None')ax1.set_xlabel('')ax1.set_xticklabels([]) #设置x轴图例为空值ax1.set_ylabel('kind')p2 = sns.heatmap(pt, linewidths = 0.05, ax=ax2, vmax=15000, vmin=0, cmap=cmap, center=None, robust=False, annot=False,mask=pt<10000 ) # mask: boolean array or DataFrameax2.set_title('mask: boolean DataFrame')ax2.set_xlabel('region')ax2.set_ylabel('kind')f.savefig('sns_heatmap_mask.jpg', bbox_inches='tight')

2.5 xticklabels, yticklabels

xticklabels: 如果是True，则绘制dataframe的列名。如果是False，则不绘制列名。如果是列表，则绘制列表中的内容作为xticklabels。如果是整数n，则绘制列名，但每个n绘制一个label。默认为True。

yticklabels: 如果是True，则绘制dataframe的行名。如果是False，则不绘制行名。如果是列表，则绘制列表中的内容作为yticklabels。如果是整数n，则绘制列名，但每个n绘制一个label。默认为True。默认为True。

f, (ax1,ax2) = plt.subplots(figsize = (10, 8),nrows=2)cmap = sns.cubehelix_palette(start = 1.5, rot = 3, gamma=0.8, as_cmap = True)p1 = sns.heatmap(pt, linewidths = 0.05,ax=ax1, vmax=15000, vmin=0, cmap=cmap, center=None, robust=False, mask=None, xticklabels=False )# robust默认为Falseax1.set_title('xticklabels=None')ax1.set_xlabel('')# ax1.set_xticklabels([]) #设置x轴图例为空值ax1.set_ylabel('kind')p2 = sns.heatmap(pt, linewidths = 0.05, ax=ax2, vmax=15000, vmin=0, cmap=cmap, center=None, robust=False, annot=False,mask=None,xticklabels=3, yticklabels=list(range(20)) ) # mask: boolean array or DataFrameax2.set_title('xticklabels=3, yticklabels is a list')ax2.set_xlabel('region')ax2.set_ylabel('kind')f.savefig('sns_heatmap_xyticklabels.jpg', bbox_inches='tight')

2.6 annot

annotate的缩写，annot默认为False，当annot为True时，在heatmap中每个方格写入数据

annot_kws，当annot为True时，可设置各个参数，包括大小，颜色，加粗，斜体字等

np.random.seed(0)x = np.random.randn(10, 10)f, (ax1, ax2) = plt.subplots(figsize=(8,8),nrows=2)sns.heatmap(x, annot=True, ax=ax1)sns.heatmap(x, annot=True, ax=ax2, annot_kws={
    'size':9,'weight':'bold', 'color':'blue'})# Keyword arguments for ax.text when annot is True.# http://stackoverflow.com/questions/35024475/seaborn-heatmap-key-wordsf.savefig('sns_heatmap_annot.jpg')

**关于annot_kws的设置，还有很多值得研究的地方，ax.text有很多属性，有兴趣的可以去研究下；

ax.text可参考官方文档：

2.7 fmt

fmt，格式设置

np.random.seed(0)x = np.random.randn(10, 10)f, (ax1, ax2) = plt.subplots(figsize=(8,8),nrows=2)sns.heatmap(x, annot=True, ax=ax1)sns.heatmap(x, annot=True, fmt='.1f', ax=ax2)f.savefig('sns_heatmap_fmt.jpg')

3 案例应用：突出显示某些数据

3.1 method 1：利用mask来实现

f,ax=plt.subplots(figsize=(10,5))x = np.random.randn(10, 10)sns.heatmap(x, annot=True, ax=ax)sns.heatmap(x, mask=x < 1, cbar=False, ax=ax,            annot=True, annot_kws={
    "weight": "bold"})f.savefig('sns_heatmap_eg1.jpg')

3.2 method 2：利用ax.texts来实现

f,ax=plt.subplots(figsize=(10,5))flights = sns.load_dataset("flights")flights = flights.pivot("month", "year", "passengers")pic = sns.heatmap(flights, annot=True, fmt="d", ax=ax)for text in pic.texts:    text.set_size(8)    if text.get_text() == '118':        text.set_size(12)        text.set_weight('bold')        text.set_style('italic')f.savefig('sns_heatmap_eg2.jpg')