CODEWITHSUNDEEP
pythonpandasranking
user27074
user27074

Reputation: 637

Panda ranking: move certain value to bottom

I want to create a weighted ranking for a data frame based on some other variables (see example below). However, a 0 means that the data for the score are missing and the row should therefore get the lowest ranking value. I know about the na_option, but I would prefer not to change the values in the Score columns. Since I want to try different boolean combinations the code has to be flexible to ascending and descending ranking. 

    Id Score1 Score2 Score3
    1    5       0      8 
    2   -4       2      6
    3    3       1      5
    4    0      -4     -3

w1, w2, w3 = 0.4, 0,3, 0.3
boolean1, boolean2, boolean3 = True, False, True
df['tmp_rank1'] = df[Score1].rank(ascending=boolean1)
df['tmp_rank2'] = df[Score2].rank(ascending=boolean2)
df['tmp_rank3'] = df[Score3].rank(ascending=boolean3)
df['final_rank'] = df['tmp_rank1'] * w1 + df['tmp_rank2'] * w2 + df['tmp_rank3'] * w3

Upvotes: 1

Views: 80

Answers (1)

Ben.T
Ben.T

Reputation: 29635

IIUC, I think you can use mask to replace 0 with nan before creating the rank and na_option, but never assign the nan back to the original column 

w1, w2, w3 = 0.4, 0.3, 0.3
boolean1, boolean2, boolean3 = True, False, True
df['tmp_rank1'] = df['Score1'].mask(df['Score1'].eq(0))\
                              .rank(ascending=boolean1, na_option='bottom')
df['tmp_rank2'] = df['Score2'].mask(df['Score2'].eq(0))\
                              .rank(ascending=boolean2, na_option='bottom')
df['tmp_rank3'] = df['Score3'].mask(df['Score3'].eq(0))\
                              .rank(ascending=boolean3, na_option='bottom')
df['final_rank'] = df['tmp_rank1'] * w1 + df['tmp_rank2'] * w2 + df['tmp_rank3'] * w3

Upvotes: 1

Related Questions