数据特征分析-分布分析

分布分析用于研究数据的分布特征，常用分析方法：

1、极差

2、频率分布

3、分组组距及组数

df = pd.DataFrame({'编码':['001','002','003','004','005','006','007','008','009','010','011','012','013','014','015'],\
                   '小区':['A村','B村','C村','D村','E村','A村','B村','C村','D村','E村','A村','B村','C村','D村','E村'],\
                   '朝向':['south','east_north','south','east_south','eath_south','north','east_north','west_north','south','west','north','east_north','south','south','east'],\
                   '单价':[7374,6435,6643,5874,6738,6453,5733,6034,5276,5999,6438,5864,6099,5699,6999],\
                   '首付':[15,7.5,18,10,30,10,18,30,40,30,20,22,29,30,40],\
                   '总价':[50,65,68,73,80,55,45,70,59,57,40,60,50,48,60],\
                   '经度':[114.0,114.6,114.8,114.2,114.5,114.3,114.4,114.7,114.9,114.1,114.8,114.2,114.5,114.3,114.8],\
                   '纬度':[22.0,22.4,22.6,22.8,22.2,22.1,22.7,22.5,22.9,22.3,22.8,22.2,22.1,22.7,22.5]    
                    }) 

 

先对总体做关于经纬度的散点图

plt.scatter(df['经度'],df['纬度'],s = df['单价']/50,c = df['总价'],cmap='Greens')   #原点的大小可以表示单价，越大单价越高；颜色深浅可以表示总价，越深总价越高

 

 求总价、单价和首付的极差

def d_range(df,*cols):
    krange = []
    for c in cols:
        crange = df[c].max() - df[c].min()
        krange.append(crange)
    return ('%s极差:%s\n%s极差:%s\n%s极差:%s'%(cols[0],krange[0],cols[1],krange[1],cols[2],krange[2]))
print(d_range(df,'总价','单价','首付'))
# 总价极差:40
# 单价极差:2098
# 首付极差:32.5

 

单价和总价的频率分布

fig,axes = plt.subplots(1,2,figsize = (10,4))
df['单价'].hist(bins = 8,ax = axes[0])
df['总价'].hist(bins = 8,ax = axes[1])

 

将总价分为8个区间，求出每个区间的频数、频率，并求出累计频率

# 频率分布，分组区间
total_range = pd.cut(df['总价'],8)   #通过cut将总价分为8个区间
total_range_count = total_range.value_counts(sort=False)   #求每个区间的个数，结果为一个Seris，不按列的大小排序
total_range_s = pd.DataFrame(total_range_count)  #将Seris转化为DataFrame，生成一个用于统计总价的DataFrame
# # total_range_s.rename(columns = {total_range_count.name:'频数',inplace = True})
total_range_s.columns = ['频数']  #给转化后的DataFrame重命名列
df['区间'] = total_range.values  #给原数据加一列区间
total_range_s['频率'] = total_range_s['频数']/total_range_s['频数'].sum()  #求总价在每个区间出现的频率
total_range_s['累计频率'] = total_range_s['频率'].cumsum()   ##求总价在每个区间的累计频率
total_range_s['频率%'] = total_range_s['频率'].apply(lambda x:'%.2f%%'%(100*x))  #格式化频率列，显示为2位百分数
total_range_s['累计频率%'] = total_range_s['累计频率'].apply(lambda x:'%.2f%%'%(100*x))#格式化频率列，显示为2位百分数
total_range_s.style.bar(subset = ['频率','累计频率'])

 

 对每个总价区间出现的频率做柱状图

total_range_s['频率'].plot(kind = 'bar',alpha = 0.8,title ='total price interval')
x = range(len(total_range_s.index))
for i,j,k in zip(x,total_range_s['频率'],total_range_s['频数']):
    plt.text(i,j+0.01,k)

 

 

对于单个字段比如朝向，做频率统计分析

# 频率分布 定性字段
cx = df['朝向'].value_counts()
cx_s = pd.DataFrame(cx)
cx_s.columns = ['频数']
cx_s['频率'] = cx_s['频数']/cx_s['频数'].sum()
cx_s['累计频率'] = cx_s['频率'].cumsum()
cx_s['频率%'] = cx_s['频率'].apply(lambda x:'%.2f%%'%(100*x))
cx_s['累计频率%'] = cx_s['累计频率'].apply(lambda x:'%.2f%%'%(100*x))
cx_s.style.bar(subset = ['频率','累计频率'] )

 

 对朝向做柱状图和饼图

fig,axes = plt.subplots(1,2,figsize = (10,4))
cx_s['频率'].plot(kind = 'bar',ax = axes[0],title = 'direction bar')   
    
plt.pie(cx_s['频数'],labels=cx_s.index,autopct='%2.f%%')
plt.title('direction pie')