我在训练我的数据集时遇到了 C50 问题。在发帖之前，我已经研究了其他人遇到过的所有类似问题和解决方案。然而，我的数据库中并没有他们遇到的问题，但仍然在 R 中执行 C50 时失败了。我的数据集看起来像这样：
'data.frame': 113967 obs. of 15 variables:$ region : Factor w/ 51 levels "US:AK","US:AL",..: 2 3 3 4 4 4 4 5 5 5 ...$ city : Factor w/ 6396 levels "179708","179720",..: 24 156 156 194 214 226 244 276 316 407 ...$ dma : Factor w/ 211 levels "1","500","501",..: 24 148 148 173 173 173 189 195 204 208 ...$ user_day : Factor w/ 7 levels "0","1","2","3",..: 6 6 6 6 6 6 6 6 6 6 ...$ user_hour : Factor w/ 24 levels "0","1","10","11",..: 5 16 16 4 22 7 10 11 15 21 ...$ os_extended : Factor w/ 71 levels "0","100","113",..: 55 68 68 7 29 14 14 14 29 34 ...$ browser : Factor w/ 19 levels "0","10","11",..: 19 18 18 8 18 9 18 17 18 18 ...$ domain : Factor w/ 2685 levels "0calc.com","100daysofrealfood.com",..: 1709 777 777 1406 727 2658 1406 1604 964 2658 ...$ position : Factor w/ 3 levels "0","1","2": 1 2 2 1 1 2 1 1 1 2 ...$ placement : Factor w/ 5406 levels "10004098","10008956",..: 3331 1696 1714 3600 438 479 3598 3423 5406 479 ...$ publisher : Factor w/ 1641 levels "1000773","1000776",..: 581 687 687 663 1369 1525 663 624 1641 1525 ...$ seller_member_id : Factor w/ 304 levels "1001","1019",..: 19 101 101 40 19 35 40 40 75 35 ...$ user_group : Factor w/ 1000 levels "0","1","10","100",..: 252 243 243 363 343 342 162 380 122 212 ...$ size : Factor w/ 7 levels "160x600","300x250",..: 5 2 2 4 5 2 2 1 2 2 ...$ predict.bid.vector.bin: Factor w/ 2 levels "(0.112,0.831]",..: 1 1 1 1 1 1 1 2 1 2 ...如您所见，最后一个变量是我的目标变量（作为因子），这里的所有特征都具有超过一个的级别。此外，数据集中没有 NA。然而，当我执行 C50 时，我得到了以下错误：
> library(C50)> myC50_Tree <- C5.0(x = test_set[,-15], y = test_set$predict.bid.vector.bin)c50 code called exit with value 1> summary(myC50_Tree)Call:C5.0.default(x = test_set[, -15], y = test_set$predict.bid.vector.bin)C5.0 [Release 2.07 GPL Edition] Fri Apr 13 14:29:54 2018-------------------------------*** line 6 of `undefined.names': attribute `region' has only one value `US'Error limit exceeded这里的问题是什么？
***您可以使用以下 R 代码获取我的模拟数据集：
# --- Set unique feature valuesregion <- c("US:AL","US:AR","US:AZ","US:CA","US:CO","US:CT","US:DC","US:FL")city <- c("179944","180802","181120","181212","181251","181315","181400","181512","181762","181842","181934","181953","182259","182295")dma <- c('522','693','754','875','345','234')user_day <- c('1','2','3','4','5','6')user_hour <- c('12','11','10','9','8','7','6','5')os_extended <- c('187','92','125','87','90')browser <- c('8','9','18','5')domain <- c('yahoo.com','youtube.com','mmctw.com','msn.com','frive.com','wework.com')position <- c('0','1','2','3')placement <- c('`234123412','34563451','235234624','46785467','234556834','85991927394')publisher <- c('5345','57867','78034','123452','84567','245645','956752')seller_memeber_id <- c('234','745','546','687','235')user_group <- c('112','556','009','345','238')size <- c('100X20','340X10','300X500','300X600')predict.bid.vector.bin <- c('(0.831,1.55]', '(0.112,0.831]')features <- list(region,city,dma,user_day,user_hour,os_extended,browser,domain,position,placement,publisher,seller_memeber_id,user_group,size,predict.bid.vector.bin)# --- Sample simulated datasettest_set <- vector()for (feature in 1:length(features)) { test_set <- cbind(test_set, sample(features[[feature]],1000,replace=TRUE))}test_set <- data.frame(test_set)colnames(test_set) <- c('region','city','dma','user_day','user_hour', 'os_extended','browser','domain','position', 'placement','publisher','seller_memeber_id', 'user_group','size','predict.bid.vector.bin')# --- check datastr(test_set)回答：
问题出在变量名 region 上 —— 我认为 C5.0 不喜欢其中的冒号。我重新创建了您的数据集，像这样：
region <- c("AL","AR","AZ","CA","CO","CT","DC","FL")然后它就正常运行了，没有错误：
treeModel <- C5.0(x=test_set[,-15],y=test_set[,15])treeModel...Evaluation on training data (1000 cases): Decision Tree ---------------- Size Errors 103 220(22.0%) << (a) (b) <-classified as ---- ---- 358 122 (a): class 1 98 422 (b): class 2Attribute usage:100.00% user_hour 28.30% region 27.30% dma 24.30% city 17.60% user_day 15.40% size 12.70% placement 9.10% user_group 7.90% browser 6.50% os_extended 4.70% publisher 4.40% position 3.70% domain 3.00% seller_memeber_id我还将因变量重新编码为 1 和 2，以防范围字符串给它带来了问题，但这似乎并不重要（然而在上面的输出中您会看到它预测为类别 1 和类别 2，这就是原因）。
