R Package for Variable Level Monitoring

An important part of model building is the “proc eyeball” sanity check. It can also be a painful part of the process, when you are the data scientist tasked with creating and checking 10,000 or more near-identical plots. The otvPlots package is designed to streamline this process. otvPlots is an R package which takes a csv file as input and provides a pdf of VLM plots and csv files of summary statistics as output, optionally ordered so that any severely abnormal time series will be at the top of the pdf. The only strict requirement of the data scientist is to specify which column of the input data file contains the date variable.

otvPlots is efficiently implemented using data.table and ggplot2 packages in R. Plots are automatically labeled if a variable dictionary is provided. Important variables can be given a highlighted label. A custom fuzzy matching algorithm can be provided by the user.

Discrete and numeric variables are handled automatically and given separate treatment. All binary variables are treated as categorical.

Output files generated by this package

A PDF file of plots, with each individual page on one variable.

For each numerical variable, the output plots include * side-by-side boxplots (left), * a trace plot of p1, p50, and p99 percentiles, * a trace plot of mean and +-1 SD control limits, and * a trace plot of missing and zero rates (bottom right).

Here is an example page of plots for a numerical variable

numerical plot

For each categorical variable (including a numerical variable with no more than 2 unique levels not including NA), the output plots include * a frequency bar plot (left), and * a grid of trace plots on categories’ proportions over time (right).

Here is an example page of plots for a categorical variable

categorical plot

CSV file(s) on summary statistics of variables, both globally and over time.

The order of variables in the CSV files is the same as in the PDF file. * A CSV file for numerical variables, including the number of observations (counts), p1, p25, p50, p75, and p99 quantiles, mean, SD, missing and zero rates. * A CSV file for categorical variables, including the number of observations (counts) and categories’ proportions. Each row is a category of a categorical (or binary) variable. The row whose category == 'NA' corresponds to missing. Categories among the same variable are ordered by global prevalence in a descending order.

Installation

Open an R (or RStudio) console and run the following code:

  1. Install the devtools package if not yet. You only need to do this once, so feel free to skip this step if the devtools is already installed. You will be asked to select a CRAN mirror.
install.packages("devtools")
  1. Install the otvPlots package

    devtools::install_github("capitalone/otvPlots")

You can also build the package yourself by cloning the repo, setting your working directory to the otvPlots folder and running devtools::build() in R, after installing the devtools package.

Note that otvPlots does depend on R and several R packages to run. You can see a complete and up to date list of dependencies in the Imports field in the DESCRIPTION file.

Getting Started

Load the package

Open an R console (or RStudio). Load the otvPlots pacakge first (all its dependent packages should be loaded automatically).

library(otvPlots)

The main function of the package is vlm. Before execute this function, input data need to be prepared using the PrepData function. Please check out the help files to see all options and many usage examples (highly recommended!)

help(vlm)
help(PrepData)

Examples

The data bankData and its labels bankLables are built-in datasets in the otvPlots package.

The first example

After running the following code, a pdf file named “bank.pdf” and two csv files named “bank_numerical_summary.csv” and “bank_categorical_summary.csv” will be generated in the current working directory.

## Load the datasets
data(bankData)
data(bankLabels)

## Prepare data and labels
bankData <- PrepData(bankData, dateNm = "date", dateGp = "months", 
                     dateGpBp = "quarters")
bankLabels <- PrepLabels(bankLabels)

## Generate a pdf file of vlm plots, and csv files of summary statistics
vlm(dataFl = bankData, dateNm = "date", labelFl = bankLabels, 
    sortFn = "OrderByR2", dateGp = "months", dateGpBp = "quarters", outFl = "bank")

More examples on the bankData data

The PrepData function only needs to be run once on a dataset. After that vlm can be run directly with the argument dataNeedPrep = FALSE (the default).

vlm(dataFl = bankData, dateNm = "date", labelFl = bankLabels, genCSV = FALSE,
    sortFn = "OrderByR2", dateGp = "months", dateGpBp = "quarters", outFl = "bank2")
bankData[, weight := rnorm(.N, 1, .1)]
bankData[, weight := weight / mean(weight)]
vlm(dataFl = bankData, dateNm = "date", labelFl = bankLabels,
    dateGp = "months", dateGpBp = "quarters", weightNm = "weight", outFl = "bank3")
sortVars <- sort(bankLabels[varCol!="date", varCol])
vlm(dataFl = bankData, dateNm = "date", labelFl = bankLabels, 
    dateGp = "months", dateGpBp = "quarters", outFl = "bank4", 
    sortVars = sortVars)
vlm(dataFl = bankData, dateNm = "date", labelFl = bankLabels, 
    dateGp = "months", dateGpBp = "quarters", outFl = "bank5", 
    varNms = "age", sortVars = NULL)

Citations

All examples for this package come from the Bank Marketing dataset available at the UCI Machine Learning Repository. The UCI repository maintains a free collection of datasets for researchers at its website.

Moro et al., S. Moro, P. Cortez, and P. Rita (2014). A Data-Driven Approach to Predict the Success of Bank Telemarketing. Decision Support Systems, Elsevier, 62:22-31, June 2014

Lichman, M. (2013). UCI Machine Learning Repository [http://archive.ics.uci.edu/ml]. Irvine, CA: University of California, School of Information and Computer Science.

Licensed under the Apache License, Version 2.0 (the “License”); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an “AS IS” BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

External Contributors

Contributors: We welcome your interest in Capital One’s Open Source Projects (the “Project”).

Any Contributor to the project must accept and sign a CLA indicating agreement to the license terms. Except for the license granted in this CLA to Capital One and to recipients of software distributed by Capital One, you reserve all right, title, and interest in and to your contributions; this CLA does not impact your rights to use your own contributions for any other purpose.

Link to Individual CLA

Link to Corporate CLA

This project adheres to the Open Source Code of Conduct. By participating, you are expected to honor this code.