machine learning - zero-inflated negative binomial regression returns singular matrix error when integrating with recursive feature elimination in R -

i'm trying use caret's recursive feature elimination on zero-inflated negative binomial regression. have created set of customized functions per github tutorial here: http://topepo.github.io/caret/rfe.html

here custom functions:

rfzeroinfl <-  list(            # explicit default 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$zero)              # find smallest coefficient either count or 0 process              #  every variable in analysis              print(paste("num coefs is: ", tostring(length(coefs.both))))              coefs_agg <- aggregate(coefs,                                     = list(names(t)),                                     fun = function(x_coefs){                                       x_out <- min(abs(x_coefs))                                       return(x_out)                                     })               colnames(coefs_agg) <- c("var", "coeffs")              coefs_sort <- coefs_agg[order(-coeffs_agg),]              print(head(coefs_sort))              coefs_sort            },            # explicit default add print statement            selectsize = function(x, metric, maximize){              print ("size function called")              best <- if (maximize)                which.max(x[, metric])              else which.min(x[, metric])              min(x[best, "variables"])            },            # explicit default add print statement            selectvar = function(y, size){              print ("var function called")              finalimp <- ddply(y[, c("overall", "var")], .(var),                                 function(x) mean(x$overall, na.rm = true))              names(finalimp)[2] <- "overall"              finalimp <- finalimp[order(finalimp$overall, decreasing = true),                ]              as.character(finalimp$var[1:size])              }) 

and here how i've called them:

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) 

here final outputs , error message:

[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 computationally singular: reciprocal condition number = 8.59945e-18" 

the output seems suggest there matrix somewhere is not invertible, can't figure out how being reached. else have experience trying put these things together? in advance suggestions or ideas.