Oracle R Enterprise comes with a vast array of features that not really documented anywhere. One of these features that I've recently found useful is the ore.recode() function.
The following code illustrates how you can records the values in an existing attributes or (more specifically in this example) how you can create a new attribute based on the values in another attribute.
The data set that I'm using is the White Wine data set that can be found on the UCI Machine Learning Repository Archive website. You can download this data set and load it into a table in your Oracle schema using just two commands.
> WhiteWine = read.table("http://archive.ics.uci.edu/ml/machine-learning-databases/wine-quality/winequality-white.csv",
sep=";", header=TRUE)
> ore.create(WhiteWine, table="WHITE_WINE")
This data set has an attribute called "quality". This "quality" attribute contains values ranging from 2 to 8, and indicates the quality of the wine.
A typical task you may want to do is to relabel values into attributes to something a bit more meaningful or to group some values into a more standardised value.
To demonstrate this I want to create a new attribute that contains a description of the type of wine (and who I might share it with).
In this case, and to allow for other values in future versions of the data sets I've coded up the following:
quality grade ------- ---------------- 1 Paint Stripper 2 Vinegar 3 Barely Drinkable 4 For the in-laws 5 For my family 6 To share with friends 7 For cooking 8 To share with my wife 9 Mine all Mine
The next step we need to perform is to gather some information about the values in the "quality" attribute. We can use the table command to quickly perform the aggregations, and then use the marplot function to graph the distributions.
> WHITE_WINE2 <- WHITE_WINE > table(WHITE_WINE2$quality) > barplot(table(WHITE_WINE2$quality), xlab="Wine Quality Ranking")
Now we are ready to perform the recoding of the values using the ore.recode() function.
> WHITE_WINE2$grade <- ore.recode(WHITE_WINE2$quality, old=c(1, 2, 3, 4, 5, 6, 7, 8, 9),
new=c("1-Paint Stripper", "2-Vinegar", "3-Barely Drinkable",
"4-For the in-laws", "5-For my family", "6-To share with friends",
"7-For cooking", "8-To share with my wife",
"9-Mine all Mine"))
You can now go and inspect the data, perform a frequency count and compare the values with what we had previously.
> head(WHITE_WINE2[,c("quality", "grade")])
> table(WHITE_WINE2$grade)
The final step is to write the newly modified data set back to your Oracle schema into a new table. This is to ensure that the original data is modified so that it can be used or reused later.
> ore.create(WHITE_WINE2, "WHITE_WINE2")
