Remove reversal transaction

Question

I have transaction level data with some reversal transactions. These transactions are denoted by a negative amount and then a counterpart positive.

trnx_df <- data.frame(Date = c("2018-01-01", "2018-01-01", "2018-01-01", "2018-01-01", "2018-01-03", "2018-01-03", "2018-01-05", "2018-02-01",
                            "2018-02-01", "2018-02-01"),
                   Product = c("A", "A", "A", "A", "B", "B", "B", "A", "A", "A"),
                   Amount = c(-1000, 1000, 1000, 1000, -1000, 1000, 500, -2000, 1000, 2000))

trnx_df

             Date Product Amount
    1  2018-01-01       A  -1000
    2  2018-01-01       A   1000
    3  2018-01-01       A   1000
    4  2018-01-01       A   1000
    5  2018-01-03       B  -1000
    6  2018-01-03       B   1000
    7  2018-01-05       B    500
    8  2018-02-01       A  -2000
    9  2018-02-01       A   1000
    10 2018-02-01       A   2000

I want to arrive at total amount and maximum amount spent by that customer on particular product.

By using dplyr I arrive at:

library(dplyr)

trnx_summary <- trnx_df %>%
group_by(Product) %>%
summarize(Total_amount = sum(Amount),
        Max_amount = max(Amount))

trnx_summary
  Product Total_amount Max_amount
1       A         3000       2000
2       B          500       1000

For total there will be no problem as the negative entry will cancel out positive one but for maximum amount spent I will get wrong output.

The maximum amount for Product A should be 1000 (2000 and -2000 will cancel each other out).

How can I fix this? Also, is there a way to delete these reversal transactions from the dataframe itself?

"reversal transactions" does this mean if there is `1000` and `-1000`, then ignore those rows? — zx8754, Jun 20 '18 at 13:30
The negative amount denote the cancelled ones but as a transaction entry both positive and negative are captured — Jayesh Jinde, Jun 20 '18 at 14:30

Lennyy · Accepted Answer · 2018-06-20T15:12:22.563

df %>% #filter the negative transactions, save in dftemp
  filter(Amount < 0) %>% 
  mutate(Amount = abs(Amount)) -> dftemp # in dftemp, negative transactions are positive to ease looking for matches

df %>%  #filter the positive transactions that do no have a negative duplicate
  filter(Amount > 0) %>% 
  anti_join(dftemp) -> dfuniques  

df %>% 
  filter(Amount > 0) %>% #filter positive transactions
  inner_join(dftemp) %>% #merge obs that are both in the original df and in dftemp 
  group_by(Date, Product, Amount) %>%  #group by date, product and amount
  slice(-1) %>% #for each date, product & amount combo, delete 1 row (which is a duplicate of one negative and one positive transaction)
  full_join(dfuniques) %>% # join the unique positive transactions (from here on, you have your desired dataframe with negative and positive transactions that cancelled each other out deleted)
  group_by(Product) %>% 
  summarise(Total_Amount = sum(Amount), Max_Amount = max(Amount))

  Product Total_Amount Max_Amount
   <fctr>        <dbl>      <dbl>
1       A         3000       1000
2       B          500        500

score 0 · Answer 2 · answered Jun 21 '18 at 11:59

Using lead and lag functions:

trnx_df %>% 
  group_by(Product, AmountAbs = abs(Amount)) %>% 
  arrange(Product, AmountAbs, Amount) %>% 
  mutate(
    remove =
      (sign(lag(Amount, default = 0)) == -1 &
           lag(AmountAbs, default = 0) == Amount) |
      ((sign(Amount)) == -1 &
         lead(AmountAbs) == AmountAbs)) %>% 
  ungroup() %>% 
  filter(!remove) %>%
  group_by(Product) %>% 
  summarise(Total_Amount = sum(Amount), Max_Amount = max(Amount))

# # A tibble: 2 x 3
# Product Total_Amount Max_Amount
#   <fct>          <dbl>      <dbl>
# 1 A               3000       1000
# 2 B                500        500

Remove reversal transaction

2 Answers2

Linked