vtreat
is a data.frame
processor/conditioner (available for R
, and for Python
) that prepares real-world data for supervised machine learning or predictive modeling in a statistically sound manner.
vtreat
takes an input data.frame
that has a specified column called “the outcome variable” (or “y”) that is the quantity to be predicted (and must not have missing values). Other input columns are possible explanatory variables (typically numeric or categorical/string-valued, these columns may have missing values) that the user later wants to use to predict “y”. In practice such an input data.frame
may not be immediately suitable for machine learning procedures that often expect only numeric explanatory variables, and may not tolerate missing values.
To solve this, vtreat
builds a transformed data.frame
where all explanatory variable columns have been transformed into a number of numeric explanatory variable columns, without missing values. The vtreat
implementation produces derived numeric columns that capture most of the information relating the explanatory columns to the specified “y” or dependent/outcome column through a number of numeric transforms (indicator variables, impact codes, prevalence codes, and more). This transformed data.frame
is suitable for a wide range of supervised learning methods from linear regression, through gradient boosted machines.
The idea is: you can take a data.frame
of messy real world data and easily, faithfully, reliably, and repeatably prepare it for machine learning using documented methods using vtreat
. Incorporating vtreat
into your machine learning workflow lets you quickly work with very diverse structured data.
In all cases (classification, regression, unsupervised, and multinomial classification) the intent is that vtreat
transforms are essentially one liners.
The preparation commands are organized as follows:
R
regression example, Python
regression example.R
classification example, Python
classification example.R
unsupervised example, Python
unsupervised example.R
multinomial classification example, Python
multinomial classification example.In all cases: variable preperation is intended to be a “one liner.”
These current revisions of the examples are designed to be small, yet complete. So as a set they have some overlap, but the user can rely mostly on a single example for a single task type.
For more detail please see here: arXiv:1611.09477 stat.AP (the documentation describes the R
version, however all of the examples can be found worked in Python
here).
vtreat
is available as an R
package, and also as a Python
/Pandas
package.
(logo: Julie Mount, source: “The Harvest” by Boris Kustodiev 1914)
Even with modern machine learning techniques (random forests, support vector machines, neural nets, gradient boosted trees, and so on) or standard statistical methods (regression, generalized regression, generalized additive models) there are common data issues that can cause modeling to fail. vtreat deals with a number of these in a principled and automated fashion.
In particular vtreat emphasizes a concept called “y-aware pre-processing” and implements:
scale
argument to vtreat::prepare()
, using some of the ideas discussed here). This allows correct/sensible application of principal component analysis pre-processing in a machine learning context.The idea is: even with a sophisticated machine learning algorithm there are many ways messy real world data can defeat the modeling process, and vtreat helps with at least ten of them. We emphasize: these problems are already in your data, you simply build better and more reliable models if you attempt to mitigate them. Automated processing is no substitute for actually looking at the data, but vtreat supplies efficient, reliable, documented, and tested implementations of many of the commonly needed transforms.
To help explain the methods we have prepared some documentation:
Data treatments are “y-aware” (use distribution relations between independent variables and the dependent variable). For binary classification use designTreatmentsC()
and for numeric regression use designTreatmentsN()
.
After the design step, prepare()
should be used as you would use model.matrix. prepare()
treated variables are all numeric and never take the value NA or +-Inf (so are very safe to use in modeling).
In application we suggest splitting your data into three sets: one for building vtreat encodings, one for training models using these encodings, and one for test and model evaluation.
The purpose of vtreat
library is to reliably prepare data for supervised machine learning. We try to leave as much as possible to the machine learning algorithms themselves, but cover most of the truly necessary typically ignored precautions. The library is designed to produce a data.frame
that is entirely numeric and takes common precautions to guard against the following real world data issues:
Categorical variables with very many levels.
We re-encode such variables as a family of indicator or dummy variables for common levels plus an additional impact code (also called “effects coded”). This allows principled use (including smoothing) of huge categorical variables (like zip-codes) when building models. This is critical for some libraries (such as randomForest
, which has hard limits on the number of allowed levels).
Rare categorical levels.
Levels that do not occur often during training tend not to have reliable effect estimates and contribute to over-fit. vtreat helps with 2 precautions in this case. First the rareLevel
argument suppresses levels with this count our below from modeling, except possibly through a grouped contribution. Also with enough data vtreat attempts to estimate out of sample performance of derived variables. Finally we suggest users reserve a portion of data for vtreat design, separate from any data used in additional training, calibration, or testing.
Novel categorical levels.
A common problem in deploying a classifier to production is: new levels (levels not seen during training) encountered during model application. We deal with this by encoding categorical variables in a possibly redundant manner: reserving a dummy variable for all levels (not the more common all but a reference level scheme). This is in fact the correct representation for regularized modeling techniques and lets us code novel levels as all dummies simultaneously zero (which is a reasonable thing to try). This encoding while limited is cheaper than the fully Bayesian solution of computing a weighted sum over previously seen levels during model application.
Missing/invalid values NA, NaN, +-Inf.
Variables with these issues are re-coded as two columns. The first column is clean copy of the variable (with missing/invalid values replaced with either zero or the grand mean, depending on the user chose of the scale
parameter). The second column is a dummy or indicator that marks if the replacement has been performed. This is simpler than imputation of missing values, and allows the downstream model to attempt to use missingness as a useful signal (which it often is in industrial data).
Extreme values.
Variables can be restricted to stay in ranges seen during training. This can defend against some run-away classifier issues during model application.
Constant and near-constant variables.
Variables that “don’t vary” or “nearly don’t vary” are suppressed.
Need for estimated single-variable model effect sizes and significances.
It is a dirty secret that even popular machine learning techniques need some variable pruning (when exposed to very wide data frames, see here and here). We make the necessary effect size estimates and significances easily available and supply initial variable pruning.
The above are all awful things that often lurk in real world data. Automating these steps ensures they are easy enough that you actually perform them and leaves the analyst time to look for additional data issues. For example this allowed us to essentially automate a number of the steps taught in chapters 4 and 6 of Practical Data Science with R (Zumel, Mount; Manning 2014) into a very short worksheet (though we think for understanding it is essential to work all the steps by hand as we did in the book). The 2nd edition of Practical Data Science with R covers using vtreat
in R
in chapter 8 “Advanced Data Preparation.”
The idea is: data.frame
s prepared with the vtreat
library are somewhat safe to train on as some precaution has been taken against all of the above issues. Also of interest are the vtreat
variable significances (help in initial variable pruning, a necessity when there are a large number of columns) and vtreat::prepare(scale=TRUE)
which re-encodes all variables into effect units making them suitable for y-aware dimension reduction (variable clustering, or principal component analysis) and for geometry sensitive machine learning techniques (k-means, knn, linear SVM, and more). You may want to do more than the vtreat
library does (such as Bayesian imputation, variable clustering, and more) but you certainly do not want to do less.
There have been a number of recent substantial improvements to the library, including:
parallel
.Some of our related articles (which should make clear some of our motivations, and design decisions):
Examples of current best practice using vtreat
(variable coding, train, test split) can be found here and here.
Trivial example:
library("vtreat")
packageVersion("vtreat")
# [1] '1.4.8'
citation('vtreat')
#
# To cite package 'vtreat' in publications use:
#
# John Mount and Nina Zumel (2019). vtreat: A Statistically Sound
# 'data.frame' Processor/Conditioner.
# https://github.com/WinVector/vtreat/,
# https://winvector.github.io/vtreat/.
#
# A BibTeX entry for LaTeX users is
#
# @Manual{,
# title = {vtreat: A Statistically Sound 'data.frame' Processor/Conditioner},
# author = {John Mount and Nina Zumel},
# year = {2019},
# note = {https://github.com/WinVector/vtreat/, https://winvector.github.io/vtreat/},
# }
# categorical example
dTrainC <- data.frame(x=c('a', 'a', 'a', 'b', 'b', NA, NA),
z=c(1, 2, 3, 4, NA, 6, NA),
y=c(FALSE, FALSE, TRUE, FALSE, TRUE, TRUE, TRUE))
dTestC <- data.frame(x=c('a', 'b', 'c', NA), z=c(10, 20, 30, NA))
# help("designTreatmentsC")
treatmentsC <- designTreatmentsC(dTrainC, colnames(dTrainC), 'y', TRUE,
verbose=FALSE)
print(treatmentsC$scoreFrame[, c('origName', 'varName', 'code', 'rsq', 'sig', 'extraModelDegrees')])
# origName varName code rsq sig extraModelDegrees
# 1 x x_catP catP 0.060049677 0.44862725 2
# 2 x x_catB catB 0.127625394 0.26932340 2
# 3 z z clean 0.237601767 0.13176020 0
# 4 z z_isBAD isBAD 0.296065432 0.09248399 0
# 5 x x_lev_NA lev 0.296065432 0.09248399 0
# 6 x x_lev_x_a lev 0.130005705 0.26490379 0
# 7 x x_lev_x_b lev 0.006067337 0.80967242 0
# help("prepare")
dTrainCTreated <- prepare(treatmentsC, dTrainC, pruneSig=1.0, scale=TRUE)
varsC <- setdiff(colnames(dTrainCTreated), 'y')
# all input variables should be mean 0
sapply(dTrainCTreated[, varsC, drop=FALSE], mean)
# x_catP x_catB z z_isBAD x_lev_NA
# 2.537498e-16 -1.268826e-16 6.336166e-17 2.536414e-16 -2.537653e-16
# x_lev_x_a x_lev_x_b
# -6.345680e-17 1.189718e-17
# all non NA slopes should be 1
sapply(varsC, function(c) { lm(paste('y', c, sep='~'),
data=dTrainCTreated)$coefficients[[2]]})
# x_catP x_catB z z_isBAD x_lev_NA x_lev_x_a
# 0.23254609 0.05841932 0.16062145 0.03162633 0.03162633 0.23254609
# x_lev_x_b
# 0.24663035
dTestCTreated <- prepare(treatmentsC, dTestC, pruneSig=c(), scale=TRUE)
print(dTestCTreated)
# x_catP x_catB z z_isBAD x_lev_NA x_lev_x_a x_lev_x_b
# 1 -1.0238626 -3.248380 7.437329 -5.420438 -5.420438 -1.0238626 0.1158472
# 2 0.7678969 -2.550396 18.374578 -5.420438 -5.420438 0.7678969 -0.2896179
# 3 3.4555361 -2.260694 29.311827 -5.420438 -5.420438 0.7678969 0.1158472
# 4 0.7678969 7.422967 0.000000 13.551095 13.551095 0.7678969 0.1158472
# numeric example
dTrainN <- data.frame(x=c('a', 'a', 'a', 'a', 'b', 'b', NA, NA),
z=c(1, 2, 3, 4, 5, NA, 7, NA), y=c(0, 0, 0, 1, 0, 1, 1, 1))
dTestN <- data.frame(x=c('a', 'b', 'c', NA), z=c(10, 20, 30, NA))
# help("designTreatmentsN")
treatmentsN = designTreatmentsN(dTrainN, colnames(dTrainN), 'y',
verbose=FALSE)
print(treatmentsN$scoreFrame[, c('origName', 'varName', 'code', 'rsq', 'sig', 'extraModelDegrees')])
# origName varName code rsq sig extraModelDegrees
# 1 x x_catP catP 2.869955e-01 0.1711651 2
# 2 x x_catN catN 2.014886e-02 0.7374034 2
# 3 x x_catD catD 3.614228e-01 0.1149323 2
# 4 z z clean 2.880952e-01 0.1701892 0
# 5 z z_isBAD isBAD 3.333333e-01 0.1339746 0
# 6 x x_lev_NA lev 3.333333e-01 0.1339746 0
# 7 x x_lev_x_a lev 2.500000e-01 0.2070312 0
# 8 x x_lev_x_b lev 1.110223e-16 1.0000000 0
dTrainNTreated <- prepare(treatmentsN, dTrainN, pruneSig=1.0, scale=TRUE)
varsN <- setdiff(colnames(dTrainNTreated), 'y')
# all input variables should be mean 0
sapply(dTrainNTreated[, varsN, drop=FALSE], mean)
# x_catP x_catN x_catD z z_isBAD
# 2.775558e-17 0.000000e+00 -2.775558e-17 4.857226e-17 6.938894e-18
# x_lev_NA x_lev_x_a x_lev_x_b
# 6.938894e-18 0.000000e+00 7.703720e-34
# all non NA slopes should be 1
sapply(varsN, function(c) { lm(paste('y', c, sep='~'),
data=dTrainNTreated)$coefficients[[2]]})
# x_catP x_catN x_catD z z_isBAD x_lev_NA x_lev_x_a
# 1 1 1 1 1 1 1
# x_lev_x_b
# 1
dTestNTreated <- prepare(treatmentsN, dTestN, pruneSig=c(), scale=TRUE)
print(dTestNTreated)
# x_catP x_catN x_catD z z_isBAD x_lev_NA x_lev_x_a
# 1 -0.250 -0.25 -0.06743804 0.9952381 -0.1666667 -0.1666667 -0.25
# 2 0.250 0.00 -0.25818161 2.5666667 -0.1666667 -0.1666667 0.25
# 3 0.625 0.00 -0.25818161 4.1380952 -0.1666667 -0.1666667 0.25
# 4 0.250 0.50 0.39305768 0.0000000 0.5000000 0.5000000 0.25
# x_lev_x_b
# 1 -2.266233e-17
# 2 6.798700e-17
# 3 -2.266233e-17
# 4 -2.266233e-17
# for large data sets you can consider designing the treatments on
# a subset like: d[sample(1:dim(d)[[1]], 1000), ]
# One can also use treatment plans as pipe targets.
dTrainN %.>%
treatmentsN %.>%
knitr::kable(.)
x_catP | x_catN | x_catD | z | z_isBAD | x_lev_NA | x_lev_x_a | x_lev_x_b | y |
---|---|---|---|---|---|---|---|---|
0.50 | -0.25 | 0.5000000 | 1.000000 | 0 | 0 | 1 | 0 | 0 |
0.50 | -0.25 | 0.5000000 | 2.000000 | 0 | 0 | 1 | 0 | 0 |
0.50 | -0.25 | 0.5000000 | 3.000000 | 0 | 0 | 1 | 0 | 0 |
0.50 | -0.25 | 0.5000000 | 4.000000 | 0 | 0 | 1 | 0 | 1 |
0.25 | 0.00 | 0.7071068 | 5.000000 | 0 | 0 | 0 | 1 | 0 |
0.25 | 0.00 | 0.7071068 | 3.666667 | 1 | 0 | 0 | 1 | 1 |
0.25 | 0.50 | 0.0000000 | 7.000000 | 0 | 1 | 0 | 0 | 1 |
0.25 | 0.50 | 0.0000000 | 3.666667 | 1 | 1 | 0 | 0 | 1 |
Related work:
Notes on controlling vtreat
’s cross-validation plans can be found here.
Note: vtreat
is meant only for “tame names”, that is: variables and column names that are also valid simple (without quotes) R
variables names.