How using function np.where along with apply lambda

Question

this code:

def nearest_independment(target):
        lst=df[df['CLINE_TYPE'].str.contains('crease') &
               df['CLINE_TYPE'].isin(['nan']).shift(2) &
               df['CLINE_TYPE'].isin(['nan']).shift(3)].index
        return min(lst, key=lambda x: abs(x - target))

df['FIGURE_TYPE'] = df.apply(lambda x:
                                       np.where(bool(re.match(r'crease',df.loc[x.name,'CLINE_TYPE']))==True &
                                                 df.loc[x.name-2,'CLINE_TYPE']=='nan' &
                                                 df.loc[x.name-3,'CLINE_TYPE']=='nan',
                                                 'INDEPENDENT',
                                        np.where((bool(re.match(r'crease',df.loc[x.name-2,'CLINE_TYPE']))==True |
                                                  bool(re.match(r'crease',df.loc[x.name-3,'CLINE_TYPE']))==True) &
                                                  bool(re.match(r'Increase',df.loc[x.name,'CLINE_TYPE']))==True &
                                                  ((df.loc[x.name,'Close_Max_FIG'] > df.loc[nearest_independment(x.name),'Open_Max_COR']) |
                                                   (df.loc[x.name,'Close_Max_FIG'] > df.loc[nearest_independment(x.name),'Close_Max_COR']))&
                                                  ((bool(re.match(r'crease',df.loc[x.name+2,'CLINE_TYPE']))==True | df.loc[x.name+2,'CLINE_TYPE']=='nan') |
                                                   (bool(re.match(r'crease',df.loc[x.name+3,'CLINE_TYPE']))==True | df.loc[x.name+3,'CLINE_TYPE']=='nan')),
                                                 'TRANZIT_UP',np.nan)),axis=1
                             )

print(df['FIGURE_TYPE'])

Problem in using df.loc, return eror 'KeyError: -2' the reason is here 'df.loc[x.name-2',logically, such a key does not exist. If you refuse to use 'loc', then the result of df['FIGURE_TYPE'] is a list of a series of values: 'nan','INDEPENDENT','TRANZIT_UP'. But, I need one value of the checked row 'lambda'. Therefore, I use 'loc'. Next, I use pd.Series for the desired result.

Data:

                CLINE_TYPE  Close_Max_FIG  Open_Max_COR  Close_Max_COR
0                      nan            NaN           NaN            NaN
1                      nan            NaN           NaN            NaN
2                      nan            NaN           NaN            NaN
3   Increase_SUPERPOSITION       29837.29      29840.65       30045.85
4                      nan            NaN           NaN            NaN
5           Decrease_FALSE       29837.29           NaN            NaN
6                      nan            NaN           NaN            NaN
7                 Increase       30045.85           NaN            NaN
8                      nan            NaN           NaN            NaN
9                      nan            NaN           NaN            NaN
10                     nan            NaN           NaN            NaN
11                     nan            NaN           NaN            NaN

Need to compare the value of rows 7 with the values ​​of row 3. Line 3 is obtained from 'nearest_independment'.

example not 'loc':

df['FIGURE_TYPE'] = df.apply(lambda x:
                                        np.where(df['CLINE_TYPE'].str.contains('crease') &
                                                 df['CLINE_TYPE'].isin(['nan']).shift(2) &
                                                 df['CLINE_TYPE'].isin(['nan']).shift(3) ,
                                                 'INDEPENDENT',
                                        np.where((df['CLINE_TYPE'].str.contains('crease').shift(2) |
                                                  df['CLINE_TYPE'].str.contains('crease').shift(3)) &
                                                  df['CLINE_TYPE'].str.contains('Increase') &
                                                  ((df.loc[x.name,'Close_Max_FIG'] > df.loc[nearest_independment(x.name),'Open_Max_COR']) |
                                                   (df.loc[x.name,'Close_Max_FIG'] > df.loc[nearest_independment(x.name),'Close_Max_COR']))&
                                                  ((df['CLINE_TYPE'].str.contains('crease').shift(-2) | df['CLINE_TYPE'].isin(['nan']).shift(-2)) |
                                                   (df['CLINE_TYPE'].str.contains('crease').shift(-3) | df['CLINE_TYPE'].isin(['nan']).shift(-3))),
                                                 'TRANZIT_UP',np.nan)),axis=1
                             )