当与R中的递归特征消除相结合时，零膨胀负二项式回归返回奇异矩阵误差

Question

我正试图在零膨胀负二项式回归上使用插入符号的递归特征消除。我根据github教程创建了一组自定义函数：http://topepo.github.io/caret/rfe.html

以下是我的自定义功能：

rfZeroinfl <-  list(
           # Explicit default to add print statement
           summary = function(data, lev=NULL, model=NULL){
             print("SUMMARY FUNCTION CALLED")
             if (is.character(data$obs))
               data$obs <- factor(data$obs, levels = lev)
             postResample(data[, "pred"], data[, "obs"])
           },
           fit = function(x, y, first, last, ...){
             print("FIT FUNCTION CALLED")
             library(pscl)
             tmp <- if (is.data.frame(x))
               x
             else as.data.frame(x)
             tmp$y <- y
             zeroinfl(y ~ ., data = tmp, dist = "negbin", EM = TRUE)
             #zeroinfl(y ~ ., data = tmp, dist = "negbin")
           },
           pred = function(object, x) {

             print("PRED FUNCTION CALLED")
             predict(object, x)
           },
           rank = function(object, x, y) {

             print("RANK FUNCTION CALLED")
             coefs <- object$coefficients
             coefs.both <- c(coefs$count, coefs$zero)
             # Find smallest coefficient for either count or zero process
             #  for every variable in the analysis
             print(paste("NUM COEFS IS: ", toString(length(coefs.both))))
             coefs_agg <- aggregate(coefs,
                                    by = list(names(t)),
                                    fun = function(x_coefs){
                                      x_out <- min(abs(x_coefs))
                                      return(x_out)
                                    })

             colnames(coefs_agg) <- c("var", "coeffs")
             coefs_sort <- coefs_agg[order(-coeffs_agg),]
             print(head(coefs_sort))
             coefs_sort
           },
           # Explicit default to add print statement
           selectSize = function(x, metric, maximize){
             print ("SIZE FUNCTION CALLED")
             best <- if (maximize)
               which.max(x[, metric])
             else which.min(x[, metric])
             min(x[best, "Variables"])
           },
           # Explicit default to add print statement
           selectVar = function(y, size){
             print ("VAR FUNCTION CALLED")
             finalImp <- ddply(y[, c("Overall", "var")], .(var), 
                               function(x) mean(x$Overall, na.rm = TRUE))
             names(finalImp)[2] <- "Overall"
             finalImp <- finalImp[order(finalImp$Overall, decreasing = TRUE),
               ]
             as.character(finalImp$var[1:size])  
           })

以下是我打电话给他们的方式：

zinb <- read.csv("http://www.ats.ucla.edu/stat/data/fish.csv")
z_x <- zinb[,c("nofish", "livebait", "camper", "persons")]
z_y <- zinb$count
z_ctrl <- rfeControl(functions = rfZeroinfl, method="repeatedcv", repeats=5, verbose=TRUE)
z_Profile <- rfe(z_x, z_y, sizes=c(3, 4), rfeControl=z_ctrl)

以下是最终输出和错误消息：

[1] "FIT FUNCTION CALLED"
+(rfe) fit Fold10.Rep5 size: 4 
[1] "FIT FUNCTION CALLED"
-(rfe) fit Fold10.Rep5 size: 4 
[1] "PRED FUNCTION CALLED"
+(rfe) imp Fold10.Rep5 
[1] "RANK FUNCTION CALLED"
[1] "NUM COEFS IS:  10"
Error in { : 
  task 1 failed - "system is computationally singular: reciprocal condition number = 8.59945e-18"

输出似乎暗示某些矩阵在某处是不可逆的，但我无法弄清楚如何达到这一点。有没有其他人有尝试将这些东西放在一起的经验？提前感谢任何建议或想法。