如何在R中将预测结果重塑为3×4格式的交叉表？

我编写了如下R代码：

v <- c("featureA", "featureB")
newdata <- unique(data[v])
print(unique(data[v])
print(predict(model, newdata, type='response', allow.new.level = TRUE)

运行后得到featureA与featureB的唯一组合：

featureA       featureB
1    bucket_in_10_to_30 bucket_in_90_to_100
2    bucket_in_10_to_30  bucket_in_50_to_90
3     bucket_in_0_to_10  bucket_in_50_to_90
4     bucket_in_0_to_10 bucket_in_90_to_100
7    bucket_in_10_to_30  bucket_in_10_to_50
10  bucket_in_30_to_100 bucket_in_90_to_100
19    bucket_in_0_to_10   bucket_in_0_to_10
33    bucket_in_0_to_10  bucket_in_10_to_50
36  bucket_in_30_to_100  bucket_in_10_to_50
38   bucket_in_10_to_30   bucket_in_0_to_10
52  bucket_in_30_to_100   bucket_in_0_to_10
150 bucket_in_30_to_100  bucket_in_50_to_90

以及对应的预测结果：

1           2           3           4           7          10          19          33          36          38          52         150 
0.001920662 0.005480186 0.000961198 0.000335883 0.006311521 0.004005570 0.000620979 0.001107773 0.013100210 0.003546136 0.007382468 0.011384935 

我希望将这些数据重塑为类似以下格式的3×4交叉表：

解决方案 #

首先修正你代码中的语法错误（缺少闭合括号）：

v <- c("featureA", "featureB")
newdata <- unique(data[v])
print(unique(data[v]))
print(predict(model, newdata, type='response', allow.new.level = TRUE))

接下来按以下步骤实现交叉表：

1. 合并特征组合与预测结果
   将特征组合数据框和预测值合并为一个完整的数据框：

   results_df <- cbind(newdata, prediction = predict(model, newdata, type='response', allow.new.level = TRUE))
   

2. 直接重塑为交叉表
   使用tidyr包的pivot_wider函数将长格式数据转为宽格式交叉表：

   library(tidyr)
   
   cross_table <- pivot_wider(results_df,
                              names_from = featureB,
                              values_from = prediction)
   

3. （可选）对featureA分组合并
   如果你需要将bucket_in_0_to_10和bucket_in_10_to_30合并为bucket_in_0_to_30，可以先重编码分组再聚合：

   library(dplyr)
   
   # 重编码featureA分组
   results_df <- results_df %>%
     mutate(featureA_group = case_when(
       featureA %in% c("bucket_in_0_to_10", "bucket_in_10_to_30") ~ "bucket_in_0_to_30",
       featureA == "bucket_in_30_to_100" ~ "bucket_in_30_to_100"
     ))
   
   # 按新分组聚合后重塑
   cross_table_grouped <- results_df %>%
     group_by(featureA_group, featureB) %>%
     summarise(prediction = mean(prediction)) # 可根据需求替换为sum等聚合方式
     %>% pivot_wider(names_from = featureB, values_from = prediction)
   

运行上述代码后，即可得到符合需求的交叉表格式数据。

内容的提问来源于stack exchange，提问作者Kyle