如何标记 DataFrame 列中 PREVIOUS 三个值相同的行？

Question

My dataframe has the following format:

final_df=pd.DataFrame([{'Open':100,'Close':101,'Candle':1},{'Open':100,'Close':101,'Candle':1},{'Open':101,'Close':102,'Candle':1},{'Open':102,'Close':101,'Candle':0},{'Open':101,'Close':100,'Candle':0},{'Open':100,'Close':99,'Candle':0},{'Open':99,'Close':98,'Candle':0},{'Open':98,'Close':99,'Candle':1},{'Open':99,'Close':100,'Candle':1},{'Open':100,'Close':101,'Candle':1},{'Open':100,'Close':99,'Candle':0}])

I would like to create a column called pattern which has the value 1 every time three or more values of the column 'candle' are the same. I tried using the following code but this code flags entries in which 3 (or more) consequent values are the same, instead I want to flag the start of this pattern, ie. have 1 for the row with the third, fourth, ... candle.

final_df['pattern'] = final_df.Candle.groupby([final_df.Candle.diff().ne(0).cumsum()]).transform('size').ge(3).astype(int) 

In this example, I want the rows with the index 2, 5, 6, and 9 to be flagged as these are the data points where the previous three rows have the same value.

Answer 1

You can try rolling the window and count the values in window are same by checking the Series.value_counts length

final_df['pattern'] = (final_df['Candle'].rolling(3).apply(lambda s: (len(s.value_counts()) == 1))
                       .fillna(0)
                       .astype(int))

print(final_df)

    Open  Close  Candle  pattern
0    100    101       1        0
1    100    101       1        0
2    101    102       1        1
3    102    101       0        0
4    101    100       0        0
5    100     99       0        1
6     99     98       0        1
7     98     99       1        0
8     99    100       1        0
9    100    101       1        1
10   100     99       0        0