可视化介绍 Matplotlib 和 Plotnine

Matplotlib 是 Python 中类似 MATLAB 的绘图工具，笔者在校学习时经常使用Matlab，所以将其视为数据可视化的入门导航，但其风格往往较规整，调整起来语法也较复杂。Plotnine则弥补了Python数据化的不足，类似R语言，可以说是ggplot2在Python上的移植版，笔者在做美赛时因其细腻的绘图风格有了初步接触，调整语法较简单，但美中不足并不支持中文，相关资料也较少。现在趁夏季学期的空档，介绍一下它们的基本用法。

构架

对于Matplotlib而言，其做的是一种串行式的加法，具有层层堆叠的特点。最底层为Canvas(画板) ，在之上构建Figure(画布) ，再在画布之上构建Axes(绘图区) ，而坐标轴(axis) 、图例(legend) 等辅助显示层以及图像层都建立在Axes之上，面向绘图对象。

这很像我们用画板作画的过程，我们先找到一块画板，然后铺上一张画布，再在画布上选定了作画的区域，在这个区域里我们画上了蓝天和白云，并在蓝天和白云旁边可以继续写字一样。

对于Plotnine而言，其做的是一种并行式的加法，具有次次递进的特点。这套图形语法把绘图过程分为了数据(data)、映射(mapping)、几何对象(geom)、统计变换(stats)、标度(scale)、坐标系(coord)、分面(facet)几个部分，各个部分是相互独立的。不同于Matplotlib，当前后之间存在较强关联时，修改往往变得繁琐，而Plotnine只需修改对应部分的设置即可。

这很像对雕像进行修饰的过程，加一个花环，戴一顶帽子，但摘下雕像上的花环和帽子也是同样的简单，特别是样式主题(theme)的选取，使得样式调整变得更加方便。

绘图函数

对于Matplotlib来说，绘图函数叫matplotlib.pyplot，一般记为plt；对于Plotnine而言，是ggplot。用好这两个基本函数，然后查找我们所需绘制的图表类型，就能完成基本的绘图任务。用例子来看，比如我们现在有这么一组数据，要去绘制柱状图：

median_age_dict = {'coutry': ['New Zealand', 'Spain', 'Ireland', 'Israel', 'Denmark', 'Norway', 'Netherlands', 'Australia', 'Italy', 'Sweden'],'age': [39.0, 37.0, 35.0, 34.0, 34.0, 34.0, 34.0, 34.0, 34.0, 34.0]}

plt

import matplotlib.pyplot as plt
import pandas as pdmedian_age_dict = {'country': ['New Zealand', 'Spain', 'Ireland', 'Israel', 'Denmark', 'Norway', 'Netherlands', 'Australia', 'Italy','Sweden'],'age': [39.0, 37.0, 35.0, 34.0, 34.0, 34.0, 34.0, 34.0, 34.0, 34.0]
}
median_age = pd.DataFrame(median_age_dict)# 创建窗体
plt.figure('matplotlib')
# 绘图
plt.bar(median_age['country'],median_age['age'])
# 展示
plt.show()

ggplot

from plotnine import *
import pandas as pdmedian_age_dict = {'country': ['New Zealand', 'Spain', 'Ireland', 'Israel', 'Denmark', 'Norway', 'Netherlands', 'Australia', 'Italy','Sweden'],'age': [39.0, 37.0, 35.0, 34.0, 34.0, 34.0, 34.0, 34.0, 34.0, 34.0]
}
median_age = pd.DataFrame(median_age_dict)# 绘图，美中不足，不支持中文
sep = (ggplot(median_age,aes('country','age',fill='country'))+# 二维条形图geom_col())
# 展示
print(sep)

看得出来，ggplot的绘图风格更为亮丽，直接展示时绘图显示要素也更为丰富，而plt则较为规整，属于加啥才能显示啥的状态，图例和坐标轴名称都没有主动显示。

风格多样化的常识

掌握柱状图、曲线图和散点图的绘制，懂得图例和坐标轴的调整，能够调整线型和点型，并能完成图片的指定形式保存，基本就能满足日常学习中的绘图需要。日后有空可再分别具体说明。

对于Matplotlib的学习， matplotlib官网上资料足够充分，民间也有很多详细的资料，如Matplotlib教程，其语法与Matlab近似，因此也可与Matlab中函数的用法类比迁移。

对于Plotnine的学习，《python数据可视化之美》提供了很好的一张导例图，特展示如下。同样可以参照R语言的ggplot2的语法进行学习，民间这方面的资料也很多，如 ggplot2高效实用指南。

Plotnine学习手册

Matplotlib模拟动态森林失火

最后展示一下自己做过的两个更为复杂的实例，第一个是模拟森林动态失火的实例，采用Matplotlib画了方格演化的动态显示。

#Here we have a 2-D battle field
#One side is green, the other is blue
import matplotlib.animation as animation  # 保存动图
import matplotlib.pyplot as plt
import numpy as np
from matplotlib.ticker import MultipleLocator
import random#define class cell
class Cell:def __init__(self,HP,GEN,site):self.HP = HP#self.POS = POSself.GEN = GENself.site = site#The LOS caused by the enemiesdef Attack(self):LOS =  random.uniform(2, 4)return LOS#The CURE provided by the teammatesdef Heal(self):CURE = random.uniform(1,3)return CURE#Judge if the cell is deaddef Sign(self):if self.HP <= 0:return Trueelse:return False#create the cell list
def create_cells(Num,HP):# the list of sitessites = []for j in range(int(np.sqrt(Num))):for k in range(int(np.sqrt(Num))):sites.append([j, k])random.shuffle(sites)# The amounts of green cellsGreen = random.randint(int(np.fix(np.sqrt(Num))), Num)cells = []for i in range(0, Green):cells.append(Cell(HP[0], 'green', sites[i]))for i in range(Green, Num):cells.append(Cell(HP[1], 'blue', sites[i]))return cells#the battle between two cells
def battle(Cell_1,Cell_2):if Cell_1.GEN == Cell_2.GEN:Cell_1.HP += Cell_2.Heal()Cell_2.HP += Cell_1.Heal()else:Cell_1.HP -= Cell_2.Attack()Cell_2.HP -= Cell_1.Attack()return [Cell_1.HP,Cell_2.HP]#the war between the cells
def war(cells):# for i in range(len(cells)):#     HP = battle(cells[i-1], cells[i])#     cells[i-1].HP = HP[0]#     cells[i].HP = HP[1]Num = len(cells)Grid = int(np.sqrt(Num))for cell_1 in cells:for cell_2 in cells:if cell_1.site[1] == cell_2.site[1]:if cell_1.site[0] == 0:if cell_2.site[0] == 1 or cell_2.site[0] == Grid:HP = battle(cell_1, cell_2)cell_1.HP = HP[0]cell_2.HP = HP[1]elif cell_1.site[0] == Grid:if cell_2.site[0] == 0 or cell_2.site[0] == Grid - 1:HP = battle(cell_1, cell_2)cell_1.HP = HP[0]cell_2.HP = HP[1]else:if cell_2.site[0] == cell_1.site[1] - 1 or cell_2.site[0] == cell_1.site[1] + 1:HP = battle(cell_1, cell_2)cell_1.HP = HP[0]cell_2.HP = HP[1]elif cell_1.site[0] == cell_2.site[0]:if cell_1.site[1] == 0:if cell_2.site[1] == 1 or cell_2.site[1] == Grid:HP = battle(cell_1, cell_2)cell_1.HP = HP[0]cell_2.HP = HP[1]elif cell_1.site[1] == Grid:if cell_2.site[1] == 0 or cell_2.site[1] == Grid - 1:HP = battle(cell_1, cell_2)cell_1.HP = HP[0]cell_2.HP = HP[1]else:if cell_2.site[1] == cell_1.site[1] - 1 or cell_2.site[1] == cell_1.site[1] + 1:HP = battle(cell_1, cell_2)cell_1.HP = HP[0]cell_2.HP = HP[1]return cells# visual the war
def visual(cells,ims):ax = plt.subplot(111)plt.ion()reddead = 0greenlive = 0bluelive = 0# get the colorfor cell in cells:x = np.linspace(cell.site[0], cell.site[0] + 1)if cell.Sign():ax.fill_between(x, cell.site[1], cell.site[1] + 1, facecolor='red')reddead += 1else:if cell.GEN == 'green':ax.fill_between(x, cell.site[1], cell.site[1] + 1, facecolor='green')greenlive += 1elif cell.GEN == 'blue':ax.fill_between(x, cell.site[1], cell.site[1] + 1, facecolor='blue')bluelive += 1# set the scaleplt.xlim(0, np.sqrt(len(cells)))plt.ylim(0, np.sqrt(len(cells)))ax.xaxis.set_major_locator(MultipleLocator(1))ax.yaxis.set_major_locator(MultipleLocator(1))ax.xaxis.grid(True, which='major')  # major,color='black'ax.yaxis.grid(True, which='major')  # major,color='black'plt.pause(0.8)ims.append(ax.findobj())# print([reddead, greenlive, bluelive])return [reddead, greenlive, bluelive]def simulata(cells):fig = plt.figure()ims = []initial = visual(cells,ims)state = [initial]i = 0while state[i][0] < initial[1] and state[i][0] < initial[2]:cells = war(cells)state.append(visual(cells, ims))i +=1ani = animation.ArtistAnimation(fig, ims, interval=500, repeat_delay=1000)ani.save("test.gif", writer='pillow')return state# Here we have a try, assume that we have Num cells
Num = 256
cells = create_cells(Num,[10,10])
state = simulata(cells)
print('Below is the result of simulation:\n',state)
greenlive_ratio = state[-1][1]/state[0][1]
bluelive_ratio = state[-1][2]/state[0][2]
print('\ngreenlive ratio is ',state[-1][1],'/',state[0][1], greenlive_ratio,'\nbluelive ratio is',state[-1][2],'/',state[0][2],bluelive_ratio)

效果显示是一个动图，并由matplotlib.animation实现动图的保存。

Plotnine点线图

另一个是用Plotnine画的一组数据的点线图，此时能看到整个风格是比较紧凑雅观的。

import numpy as np
from plotnine import *
import pandas as pddef pit():# 数据df = {'x': [], 'y': [], 'label': []}y1 = [25978.40, 25978.40, 25978.40, 25978.40, 25978.40, 25978.40,25978.40, 25978.40, 25978.40, 25978.40, 25978.40]y2 = [12511.43, 12511.43, 12511.43, 12511.43, 12511.43, 11440.20,10368.96, 14418.96, 18468.96, 24969.96, 31470.96]y3 = [47111.58, 47111.58, 47111.58, 46111.58, 45111.58, 42602.52,40093.45, 35093.45, 30093.45, 27264.94, 24436.42]y4 = [7055.95, 7055.95, 7055.95, 7555.95, 8055.95, 9550.86,11045.76, 8074.72, 5103.67, 4645.83, 4187.98]y5 = [86468.18, 86468.18, 86468.18, 86968.18, 87468.18, 89553.58,91638.97, 95560.02, 99481.06, 96266.42, 93051.78]Num = 11for i in range(Num):df['x'].append(i + 1)df['y'].append(y1[i])df['label'].append('CO')for i in range(Num):df['x'].append(i + 1)df['y'].append(y2[i])df['label'].append('WY')for i in range(Num):df['x'].append(i + 1)df['y'].append(y3[i])df['label'].append('AZ')for i in range(Num):df['x'].append(i + 1)df['y'].append(y4[i])df['label'].append('NM')for i in range(Num):df['x'].append(i + 1)df['y'].append(y5[i])df['label'].append('CA')cases = ['(0,1)', '(0.2,0.8)', '(0.4,0.6)', '(0.6,0.4)', '(0.8,0.2)', '(1,0)']df = pd.DataFrame(df)COLOR = ('#F91400', '#FF328E')SHAPE = ('o', 's')# 绘图first_plot = (ggplot(df, aes('x', 'y', shape='label', fill='label', color='label')) +# 点线图结合 geom_point(size=3) +geom_line(size=1) +labs(x=' ', y="", title=' ') +# 坐标轴调整scale_x_continuous(limits=(0, 11), breaks=np.linspace(1, 11, 6), labels=cases) +# 风格调整，这是很方便灵活的一步theme_bw())first_plot.save('alplan.png', width=10, height=4.5)print(first_plot)pit()