[R] How to use R to perform prediction based on history data
ggrothendieck at gmail.com
Tue Aug 18 15:38:56 CEST 2009
Please read the last line to every message on r-help and note the
request to provide reproducible code.
Anyways, try this:
Lines <- "Field1 Field2 Field3 Field4 ResultField
1231 CA TRUE 443 TRUE
23231 NC TRUE 123 FALSE
1231 CA FALSE 243 TRUE
23231 NC TRUE 123 NA"
DF <- read.table(textConnection(Lines), header = TRUE)
# logistic regression using first 3 rows
mod <- glm(ResultField ~., DF[1:3, ], family = binomial)
# prediction using 4th row
predict(mod, DF[4, 1:4], type = "response")
and also have a look at the caret package.
On Fri, Aug 14, 2009 at 10:27 PM, Jack Su<jacksuyu at gmail.com> wrote:
> Say I have a csv file, each row contains several fields, one of them
> are whether the row is success.
> In history data, I have all the fields including the result of whether
> it is success. In future data, I only have fields without the result.
> For example:
> history data:
> Field1 Field2 Field3 Field4 ResultField
> 1231 CA TRUE 443 TRUE
> 23231 NC TRUE 123 FALSE
> 1231 CA FALSE 243 TRUE
> The future data:
> Field1 Field2 Field3 Field4
> 23231 NC TRUE 123
> I am newbie in R and statistics, I just feel R could have some
> mechanism to give the probably of success rate based on history data.
> I tried to read in the csv data, and try to call "factor" on the list,
> but I am seeing error message:
> Error in sort.list(unique.default(x), na.last = TRUE) :
> Any idea are highly welcome.
> Thanks in advance.
> R-help at r-project.org mailing list
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
More information about the R-help