确定哪个组具有事件序列，同时忽略重复日期

如何解决确定哪个组具有事件序列，同时忽略重复日期

我已经使用 SPADE 算法确定了医疗记录（> 20 00 万条记录）中的频繁事件序列。现在我想确定哪个 ID （组）具有序列。该序列必须在 182 天内发生。同一天发生的任何事件（即 items == R-S,bxR01,S-O）不应被视为一个序列（它具有随机顺序）。

开始时，我有一个很长的 data.table，每个 ID 包含多行，每天一行在 items 列中包含一个或多个 items（请参阅下面的可重现示例）。

获取序列的一种方法似乎是拆分包含多个事件的 items 并为每 182 天的时间段制作行（使 long-data.table 更长），然后制作宽 {{1} }} 为 182 期间内的每个项目（每个 data.table 在同一天有不同的行用于唯一的 items）。

我不知道从哪里开始，所以我试图以我能想到的小步骤减少所有事情，但我陷入了下面描述的兔子洞...... 我确实认为来自 this thread 的第一条评论可能会有所帮助（但我不确定如何）。

我管理的是通过以下方式选择包含序列中任何 ID 的所有行（即，从中创建一个新的 data.table）：

items

然后我可以像这样使用 unnest：

DT2 <- DT[grepl("bxD01",items) | grepl("S-In",items)]

删除所有不相关的项目：

DT2 = DT2 %>% 
    mutate(items = strsplit(as.character(items),",")) %>% 
    unnest(items)

然后是 DT2 <- DT[grepl("bxD01",items)] 和 reorder：

aggregate

然而，这（除了非常低效之外）不会为重复的日子制作不同的行......为此我想我会使用上面提到的胎面。

样本（模拟）数据

DT2 = DT2[order(ID,days)]
# aggregate to all items in one row/cell
DT3 = as.data.table(aggregate(as.formula(items~ID),data=DT2,FUN = toString))

样本（模拟）序列列表

DT <-    structure(list(items = c("bxM01,T-Other","bxD01","S-In","bxD02","L-I","A9","R-S,bxR03","bxA02","HDTR","HVAL","SC.R","bxD11,S-Other","K-Other","bxD06,S-In","A-s.spec","LON","bxJ01","S-Other","HVAL,SC.R","bxN02,bxN02,bxC07,"bxC09","R-all-rhin","S-S,bxD07,bxD01","bxD07,ECZM","X-resp-prev","bxD07","T-Other","bxA11","P-S","bxN01,NKSH","D-S","FYS,B-Other","bxC07,RON,"bxM01,bxA01","bxS01","NKSH","bxC08","bxD04,K-Other","bxN02",Y-S",bxR06,ALGY","SC.R,"bxD10","Y-S","S-S","bxR03,LON","RON,S-Ne","Z-S","SX","F-In,bxS01","N-Other",TROT","S-Sdf","OBES","bxJ01,"bxR01,LAB,"A-Unknown","K-S",bxA02","L-Other","X-Other","bxN05","bxR06",bxA02,bxN02","TROT"),days = c(613L,861L,883L,1210L,1408L,1699L,391L,409L,745L,1448L,28L,32L,43L,98L,105L,231L,439L,442L,446L,544L,704L,801L,845L,846L,851L,1097L,1131L,1168L,1246L,1264L,1309L,1313L,1323L,1327L,1452L,1475L,1482L,1484L,1518L,1588L,1629L,1630L,1631L,1634L,1641L,1645L,1727L,1741L,1769L,1809L,790L,953L,999L,1004L,1013L,1015L,1034L,1055L,1190L,1211L,1375L,1544L,1802L,241L,353L,416L,437L,451L,547L,548L,706L,831L,832L,839L,1099L,1276L,1301L,1567L,1598L,287L,574L,854L,872L,943L,1089L,1147L,1170L,1177L,1201L,1202L,1512L,20L,30L,52L,53L,87L,309L),ID = c("G","G","F","E","D","C","B","A","A")),row.names = c(NA,-100L
),class = c("data.table","data.frame"))

总而言之，我需要一个包含二进制表达式的 data.table，该表达式指示任何 seq <-structure(list(sequence = c("<{bxD01},{S-In}>","<{L-s.spec},TROT}>","<{NKSH},{T-Other}>","<{ABDO},{SC.REF}>",{RON}>",{R-S}>","<{P-S},{P-S}>",{NKSH}>")),-45L),"data.frame")) 在 182 天内是否具有序列。

**** 编辑 **** 我忘记添加我必须在 data.table 中选择序列的代码：

ID

非常欢迎任何帮助:)

---

不太清楚问题在问什么，因此建议进行此编辑。以下是否举例说明了您正在寻找的内容？ 谢谢，是的，这就是我要问的。我也posted another question (and answer to part of this)澄清，但你的编辑可能会更清楚。

# reg expression for matching based on order
seq[,sequence_list := paste0(".*",gsub("[,]","\\s*(.*?)\\",gsub("[</{/}>]","",as.character(sequence))),".*")]

# loop for every item in the sequence list (regex expression) and then label it as the sequence result
for (NUM in 1:length(seq$sequence_list[1])) {
  DT3[grepl(seq$sequence_list[1],x = items),RES_GROUP := seq$sequence[1]]
 DT3[grepl(seq$sequence_list[1],RES_GROUP_label := 1]
}