How to coerce a list object to type 'integer', Error in R

Question

I am newbies in using R for Data Mining and Machine Learning. While I am studying about Naive Bayes Classified, I come across this error:

"Error in which((sapply(newdata[ind_factor], nlevels) != sapply(tables[ind_factor], : (list) object cannot be coerced to type 'integer'"

This is my code:

data <- read.csv(file.choose(),header = T) 
str(data)
set.seed(1234)
splitData <- sample(2,nrow(data),replace = T,prob = c(0.8,0.2))
train<-data[splitData == 1,]
test <- data[splitData == 2,]
mdl <- naive_bayes(admit ~ .,data = train)
predicted <- predict(mdl, train, type = 'prob')

When I run the final line, it throws the error message above. Can anyone help me please! Thanks a lot.

Answer 1

It looks like one of your independent variables is a string or factor variable, and all need to be numeric. See my toy dataset below. I get the same error when including all the variables; however, when I take var4 out (where variables are strings), it works).

If you want to use the variable, you could convert the string variable to a factor, then convert to the factor to a numeric variable (which will capture the underlying values of the factor).

library(naivebayes)
#data <- read.csv(file.choose(),header = T) 
data <- data.frame(admit = sample(100, x=c(F,T), prob=c(.5,.5), replace=T),
           var1 = sample(100, x=1:4, replace=T),
           var2 = sample(100, x=1:3, replace=T),
           var3 = sample(100, x=1:3, replace=T),
           var4 = sample(100, x=c("s1", "s2"), replace=T))

str(data)
set.seed(1234)
splitData <- sample(2,nrow(data),replace = T,prob = c(0.8,0.2))
train<-data[splitData == 1,]
test <- data[splitData == 2,]

# Doesn't work
mdl <- naive_bayes(admit ~ .,data = train)
predicted <- predict(mdl, train, type = 'prob')

# Works
mdl <- naive_bayes(admit ~ var1 + var2 + var3,data = train)
predicted <- predict(mdl, train, type = 'prob')

# Convert string to factor then numeric
train$var4 <- as.numeric(as.factor(train$var4))

mdl <- naive_bayes(admit ~ .,data = train)
predicted <- predict(mdl, train, type = 'prob')