statcompute / yager Goto Github PK

Hi,

I was testing your package with a sample dataset.
I'm using a training set/dataframe called train with roughly 52k observations and 5 variables, one being the target (52000x5). The test set, test, only contains 48 observations and the same 5 variables (48x5).

Since I want to predict the test target value (48 observations) directly, I was not able to use the grnn package on its own, since it only assumes one observation at a time (at least I was not able to predict it as a set of 48 observations - matrix dimensions errors). Hence, I have 3 options:

1. Predict each one of the 48 observations with a for loop
2. Predict each one of the 48 observations with a for loop, but using paralelism via %dopar% of the doParallel package.
3. Predict the whole set using grnn_parpred, since it accepts the whole set as an input.

When comparing the processing times of each of the stated option, I get the following results:

1. For Loop = ~27 seconds
2. For Loop with doParallel = ~9 seconds
3. Using grnn_parpred = ~130 seconds

Using the regular for loop as a reference, the doParallel package solves the problem much much quickly than grnn_parpred, even by predicting each observation individually.

The same mistake. What's wrong?
> install.packages("C:/Users/User/Downloads/yager_0.1.0.tar.gz", repos = NULL, type = "source") Installing package into ‘C:/Users/User/Documents/R/win-library/3.6’ (as ‘lib’ is unspecified) Ошибка в getOctD(x, offset, len) :invalid octal digit Warning in install.packages : installation of package ‘C:/Users/User/Downloads/yager_0.1.0.tar.gz’ had non-zero exit status

I am a scoreacard expert. Can I see a simple credit scoring example? Using German Credit data?