6. 高亮每列的最大值

 In[61]: pd.options.display.max_rows = 8
#  读取college数据集，INSTNM作为列
 In[62]: college = pd.read_csv('data/college.csv', index_col='INSTNM')
         college.dtypes
Out[62]: CITY                   object
         STABBR                 object
         HBCU                  float64
         MENONLY               float64
                                   ...   
         PCTFLOAN              float64
         UG25ABV               float64
         MD_EARN_WNE_P10        object
         GRAD_DEBT_MDN_SUPP     object
         Length: 26, dtype: object

#  MD_EARN_WNE_P10 和 GRAD_DEBT_MDN_SUPP 两列是对象类型，对其进行检查，发现含有字符串
 In[63]: college.MD_EARN_WNE_P10.iloc[0]
Out[63]: '30300'
 In[64]: college.MD_EARN_WNE_P10.iloc[0]
Out[64]: '30300'

#  降序检查
 In[65]: college.MD_EARN_WNE_P10.sort_values(ascending=False).head()
Out[65]: INSTNM
         Sharon Regional Health System School of Nursing    PrivacySuppressed
         Northcoast Medical Training Academy                PrivacySuppressed
         Success Schools                                    PrivacySuppressed
         Louisiana Culinary Institute                       PrivacySuppressed
         Bais Medrash Toras Chesed                          PrivacySuppressed
         Name: MD_EARN_WNE_P10, dtype: object

#  可以用to_numeric，将某列的值做强制转换
 In[66]: cols = ['MD_EARN_WNE_P10', 'GRAD_DEBT_MDN_SUPP']
         for col in cols:
             college[col] = pd.to_numeric(college[col], errors='coerce')
         college.dtypes.loc[cols]
Out[66]: MD_EARN_WNE_P10       float64
         GRAD_DEBT_MDN_SUPP    float64
         dtype: object

#  用select_dtypes方法过滤出数值列
 In[67]: college_n = college.select_dtypes(include=[np.number])
         college_n.head() 
Out[67]:

#  有的列只含有两个值，用nunique()方法挑出这些列
 In[68]: criteria = college_n.nunique() == 2
         criteria.head()
Out[68]: HBCU          True
         MENONLY       True
         WOMENONLY     True
         RELAFFIL      True
         SATVRMID     False
         dtype: bool

#  将布尔Series传给索引运算符，生成二元列的列表
 In[69]: binary_cols = college_n.columns[criteria].tolist()
         binary_cols
Out[69]: ['HBCU', 'MENONLY', 'WOMENONLY', 'RELAFFIL', 'DISTANCEONLY', 'CURROPER']

#  用drop方法删除这些列
 In[70]: college_n2 = college_n.drop(labels=binary_cols, axis='columns')
         college_n2.head()
Out[70]:

#  用idxmax方法选出每列最大值的行索引标签
 In[71]: max_cols = college_n2.idxmax()
         max_cols
Out[71]: SATVRMID                      California Institute of Technology
         SATMTMID                      California Institute of Technology
         UGDS                               University of Phoenix-Arizona
         UGDS_WHITE                Mr Leon's School of Hair Design-Moscow
                                                  ...                    
         PCTFLOAN                                  ABC Beauty College Inc
         UG25ABV                           Dongguk University-Los Angeles
         MD_EARN_WNE_P10                     Medical College of Wisconsin
         GRAD_DEBT_MDN_SUPP    Southwest University of Visual Arts-Tucson
         Length: 18, dtype: object

#  用unique()方法选出所有不重复的列名
 In[72]: unique_max_cols = max_cols.unique()
         unique_max_cols[:5]
Out[72]: array(['California Institute of Technology',
                'University of Phoenix-Arizona',
                "Mr Leon's School of Hair Design-Moscow",
                'Velvatex College of Beauty Culture',
                'Thunderbird School of Global Management'], dtype=object)

#  用max_cols选出只包含最大值的行，用style的highlight_max()高亮
 In[73]: college_n2.loc[unique_max_cols].style.highlight_max()
Out[73]:

更多

#  用axis参数可以高亮每行的最大值
 In[74]: college = pd.read_csv('data/college.csv', index_col='INSTNM')
         college_ugds = college.filter(like='UGDS_').head()
         college_ugds.style.highlight_max(axis='columns')
Out[74]:

 In[75]: pd.Timedelta(1, unit='Y')
Out[75]: Timedelta('365 days 05:49:12')