基于R中数据帧的另一个变量计算和存储一个变量的出现次数

Question

Date         HomeTeam      AwayTeam        FTHG  FTAG FTR  
   <chr>      <chr>        <chr>          <dbl> <dbl> <chr>
 1 08/10/2018 Man United   Leicester          2     1 H    
 2 08/11/2018 Bournemouth  Cardiff            2     0 H    
 3 08/11/2018 Fulham       Crystal Palace     0     2 A    
 4 08/11/2018 Huddersfield Chelsea            0     3 A    
 5 08/11/2018 Newcastle    Tottenham          1     2 A    
 6 08/11/2018 Watford      Brighton           2     0 H    
 7 08/11/2018 Wolves       Everton            2     2 D    
 8 08/12/2018 Arsenal      Man City           0     2 A    
 9 08/12/2018 Liverpool    West Ham           4     0 H  

我正在尝试动态计算每个团队在 FTR 列（'H'、'A'、'D'）中可能出现的每个值的出现次数。目标是确定每支球队最近 10 场比赛的记录。联赛数据按特定日期过滤，因此球队可能在请求的日期没有打过 10 场比赛。然后，此信息（团队的 Last10 记录）将合并到包含团队统计数据的更大表格中。因此，我认为解决此问题的最简单方法是创建一个数据框/小标题，其中团队的名称在一列中，他们的 Last10 记录在第二列中。

我见过几个连续计数和按组计数的解决方案，但没有一个是迭代地执行它，有限制，并按我的需要存储它。结果应如下所示。

1 Man United 6-1-3
2 Leicester  2-3-5
3 Arsenal    4-5-1
4 Man City   8-1-1

Answer 1

假设你的数据库在dfdata.frame，前面的代码产生一个data.frame，有四列：Teams，H，A和D，最后的结果匹配：

# Store the last 10 matches result of each team
# Check if there is a nicer way to obtain the names of all the teams
Last10 <- data.frame(Teams = names(table(rbind(df$HomeTeam, 
                                               df$AwayTeam))))
# Create empty columns to store results
Last10[,ncol(Last10)+c(1,2,3)] <- 0 
names(Last10) <- c("Teams","H", "D", "A")

MaxMatches <- 10 # Maximum number of last matches that would score
FilterDate <- df$Date[1:9] # Select which dates are taken into account
for (i in rev(which(df$Date %in% FilterDate))) { 
  # Go from the end date to the beginning
  # Check which teams have played
  teamindex <- which(Last10$Teams %in% df[i,c("HomeTeam","AwayTeam")])
  # Check if any of the teams have all the needed matches 
  teamscore <- rowSums(Last10[teamindex,c("H","D","A")])
  # If any teamscore is greater than the maximum number of matches
  # the scores should not be updated
  Last10$H[teamindex] <- Last10$H[teamindex] + 
    ifelse(teamscore < MaxMatches & df[i,"FTR"] == "H", 1, 0) 
  Last10$A[teamindex] <- Last10$A[teamindex] + 
    ifelse(teamscore < MaxMatches & df[i,"FTR"] == "A", 1, 0) 
  Last10$D[teamindex] <- Last10$D[teamindex] + 
    ifelse(teamscore < MaxMatches & df[i,"FTR"] == "D", 1, 0) 
}

如果您希望粘贴在一起的三列看起来像H-A-D，那么您可以使用：

Last10$results <- apply(Last10[,c("H","D","A")], 1, paste, collapse = "-")

Answer 2

也许以下内容可以满足问题的要求。
它使用xtabs 将感兴趣的列制成表格，然后将结果转换为类矩阵。

df2 <- reshape2::melt(df1[c(2, 3, 6)], id.vars = "FTR")[-2]
res <- xtabs(~ value + FTR, df2)
dimnames(res) <- unname(dimnames(res))
rm(df2)

head(res)
#               A D H
#Arsenal        1 0 0
#Bournemouth    0 0 1
#Brighton       0 0 1
#Cardiff        0 0 1
#Chelsea        1 0 0
#Crystal Palace 1 0 0

要将团队名称作为第一列的 data.frame，只需 cbind 行名称和矩阵。

res <- cbind.data.frame(Team_names = rownames(res), res)
rownames(res) <- NULL
head(res)
#      Team_names A D H
#1        Arsenal 1 0 0
#2    Bournemouth 0 0 1
#3       Brighton 0 0 1
#4        Cardiff 0 0 1
#5        Chelsea 1 0 0
#6 Crystal Palace 1 0 0