使用R碱基

sapply(split(df$Predicted_value,cumsum(c(1,diff(df$day_number)!=1))), sum)
   1     2     3 
0.003 0.003 0.008

使用this post的答案：

df %>%
  group_by(transaction_ID) %>%
  mutate(id = cumsum(c(1, diff(day_number) != 1))) %>%
  group_by(transaction_ID, id) %>%
  summarise(result=sum(Predicted_value))%>%
  ungroup

  transaction_ID    id result
           <int> <dbl>  <dbl>
1             12     1  0.003
2             12     2  0.003
3             13     1  0.008

根据所需的输出，它不是累计总和，而是按事务ID和日期组的总和。
使用data.table

dat = data.table(transID = c(12,...),
                 dayNum = c(1,2,...),
                 predVal = c(0.001, 0.002, ...))

# introduce a grouping column; each group starts when day == 1
dat[, 
    gr := cumsum(dayNum == 1)]

# aggregate
dat[,
    sum(predVal),
    by = gr]

    gr    V1
1:  1 0.003
2:  2 0.003
3:  3 0.008