fastknn for regression

[hutaohutc]

Hi , I want to ask can this fastknn funcion use for regression?

[davpinto]

I'm sorry but for now, unfortunately, fastknn works only for classification problems. I'm planning to include regression capabilities, but it can take a few weeks or months. Here it follows the to do list:

• [x] implement the knnExtract() method for feature extraction and dimensionality reduction
• [ ] add an automatic feature ranking procedure to the knnDecision() method
• [ ] implement the knnStack() method to generate metafeatures from the original data that can be used in stacked models
• [ ] implement the knnImpute() method for missing values imputation
• [ ] extend the fastknn() method to work with regression problems

However, some methods like knnDecision() and knnExtract() cannot be extended to deal with regression problems. They require nominal target variables.

[yunxileo]

when i run : knnDecision(x.tr, y.tr, x.te, y.te, k = 15) . Error in viridisLite::viridis(256, alpha, begin, end, direction, option) : unused argument (option)

[davpinto]

Hi @yunxileo,

I can't reproduce your problem. I need more details about your data. Have you tried the toy example in the documentation?

library(fastknn)

## Load toy data
data("spirals", package = "fastknn")

## Split data for training and test
set.seed(123)
tr.idx <- which(caTools::sample.split(Y = spirals$y, SplitRatio = 0.7))
x.tr   <- spirals$x[tr.idx, ]
x.te   <- spirals$x[-tr.idx, ]
y.tr   <- spirals$y[tr.idx]
y.te   <- spirals$y[-tr.idx]

## Plot decision boundary
knnDecision(x.tr, y.tr, x.te, y.te, k = 15)

Try it and let me know if the problem persists. In addition, check your version of the viridis package. It must be greater than 0.3.4.

[yunxileo]

I am sorry ,but the problem persists . my Info: R version 3.3.2 (2016-10-31) Platform: x86_64-w64-mingw32/x64 (64-bit) Running under: Windows 7 x64 (build 7601) Service Pack 1 and viridis_0.4.0

[davpinto]

Hi @yunxileo,

I'm not able to reproduce this problem on Ubuntu 16.04 x64, so I used devtools::build_win() to submit the package to winbuilder. Again, I'm not able to reproduce your problem. Here is the link to the report: test fastknn on Windows. It will expire in 3 days.

Note that all tests returned without error, including the tests for the knnDecision() method.

Unfortunately, I have no idea about the origin of your problem. Try to reinstall the viridis package.

If there is something else I can do to help you, please ask me! I hope you find a solution.

[yunxileo]

Actually,after I reinstall the viridisLite package,the knnDecision() works.Thanks a lot for your help!