Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Nidhi Mehta
authored and
Nidhi Mehta
committed
Nov 24, 2015
1 parent
578eb95
commit a06b9af
Showing
2 changed files
with
91 additions
and
0 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,37 @@ | ||
#This tests quantile and weighted quantile on synthetic data by comparing with R | ||
|
||
test.quantile <- function(conn){ | ||
N = 1000 | ||
|
||
x = rgamma(N, shape=0.067, scale = 0.008) | ||
aa = as.h2o(x) | ||
r_q = quantile(x, probs = c(0.1, 0.5, 1, 2, 5, 10, 50,88.83,99,90)/100,na.rm=T) | ||
h_q = h2o.quantile(aa,probs = c(0.1, 0.5, 1, 2, 5, 10, 50,88.83,99,90 )/100,na.rm=T) | ||
expect_equal(r_q,h_q ) | ||
|
||
x = rlnorm(N,meanlog = 12,sdlog = 132) | ||
aa = as.h2o(x) | ||
r_q = quantile(x, probs = seq(0,1,.05),na.rm=T) | ||
h_q = h2o.quantile(aa,probs = seq(0,1,.05),na.rm=T) | ||
expect_equal(r_q,h_q ) | ||
|
||
x = rexp(N, rate = 12.3) | ||
ss = sample(1:N,size = N/10,replace = F) | ||
x[ss]=NA | ||
aa = as.h2o(x) | ||
r_q = quantile(x, probs = seq(0,1,.05),na.rm=T) | ||
h_q = h2o.quantile(aa,probs = seq(0,1,.05),na.rm=T) | ||
expect_equal(r_q,h_q ) | ||
|
||
#weighted quantiles | ||
#library(Hmisc) | ||
x = runif(N) | ||
aa = as.h2o(x) | ||
wts = sample(1:6, N, TRUE) | ||
h_wts = as.h2o(wts) | ||
#r_q = wtd.quantile(x, wts, probs = seq(0,1,.05)) | ||
#h_q = h2o.quantile(aa,probs = seq(0,1,.05),weight_column = h_wts) | ||
#expect_equal(r_q,h_q ) | ||
|
||
} | ||
doTest("Test quantile",test.quantile ) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,54 @@ | ||
# This tests weighted quantile | ||
# by comparing results with R's wtd.quntile function and sanity checking by ignoring rows with zero weight | ||
# dataset - http://mlr.cs.umass.edu/ml/datasets/Bank+Marketing | ||
|
||
test.wtd.quantile <- function(conn){ | ||
|
||
a= h2o.importFile(locate("smalldata/gbm_test/bank-full.csv.zip"),destination_frame = "bank_UCI") | ||
dim(a) | ||
myX = 1:16 | ||
myY = 17 | ||
|
||
rowss =45211 | ||
#Sample rows for 2-fold xval | ||
ss = sample(1:rowss,size = 22000) | ||
ww = rep(1,rowss) | ||
ww[ss]=2 | ||
|
||
#Parse fold column to h2O | ||
wei = as.h2o(ww,destination_frame = "weight") | ||
colnames(wei) | ||
|
||
#Cbind fold column to the original dataset | ||
a = h2o.assign(h2o.cbind(a,wei),key = "bank") | ||
dim(a) | ||
|
||
#Build gbm by specifying the fold column | ||
gg = h2o.gbm(x = myX,y = myY,training_frame = a,ntrees = 5,fold_column = "x",keep_cross_validation_predictions = T,model_id = "cv_gbm") | ||
|
||
#Define and use weights column | ||
ww[ss]=0 | ||
wi = as.h2o(ww,destination_frame = "weight_col") | ||
|
||
#Predict | ||
pr = h2o.predict(gg,a) | ||
pred = as.data.frame(pr[,3]) | ||
|
||
# weighted h2o quantile | ||
#hq = as.numeric(h2o.quantile(pr[,3],probs = seq(0,.95,.05),weight_column = wi)) | ||
|
||
# weighted R quantile | ||
#library(Hmisc) | ||
#wq = as.numeric(wtd.quantile(pred[,1],ww,probs = seq(0,.95,.05))) | ||
|
||
#expect_equal(wq,hq,tolerance = 1e-5) | ||
|
||
#Sanity check with just nonzero weighted rows | ||
#pp=pred[which(ww==1),] | ||
#qq = as.numeric(quantile(pp,probs = seq(0,.95,.05))) | ||
|
||
#expect_equal(wq,qq,tolerance = 3e-4) | ||
#expect_equal(hq,qq,tolerance = 3e-4) | ||
} | ||
doTest("Test weighted quantile",test.wtd.quntile ) | ||
|