clarkjoe / overtime

R package to generate default summary statistic columns

overtime

Installation

devtools::install_github("clarkjoe/overtime", dependencies = TRUE)

Getting Started

The intent of overtime is to help machine learning developers generate lots of summary statistics extremely quickly. While there are many default summary statistics initially processed, including more will be available in future releases. The default summary statistics are:

1. sum
2. mean
3. median
4. sd
5. max
6. min
7. sd / mean (coefficient of variation)
8. OO2
9. OO3
10. Largest positive sequence
11. Largest negative sequence
12. Largest zero sequence
13. Largest increasing sequence
14. Largest decreasing sequence
15. Largest increasing positive sequence
16. Largest decreasing positive sequence
17. Largest increasing negative sequence
18. Largest decreasing negative sequence

The actual content of the package is smaller than most, but its scope of usability is a wide net. Below is a simple example:

library(overtime)
library(tidyverse)
library(magrittr)

data <- readRDS('../data/rawData.rds')

nestedData <- data %>%
  overtime_by("day") %>%
  overtime_get()

unnestedData <- nestedData %>%
  overtime_unnest()

BOOM! It's that simple.

---

What it looks like

nestedData

AccountNumber	D_Cognostics
A	tibble [1 x 18]
B	tibble [1 x 18]
C	tibble [1 x 18]

unnestedData

AccountNumber	D_Count	D_Count	D_Count	D_Mean	D_Median	D_SD	D_Max	D_Min	...
A	13	13	13	4.33	5	3.06	7	1	...
B	10	10	10	3.33	4	2.08	5	1	...
C	0	0	0	0	0	0	0	0	...

---

Compatible data format

Currently, only a specific data format works with overtime. Here is an example data format:

AccountNumber	Date	Count
A	2014-11-01	1
A	2014-11-02	4
B	2014-11-01	0
B	2014-11-01	12
C	2014-11-02	8
C	2014-11-01	47

There must be:

• Grouped variable
• Continious dates
  • All group variables must have equal number of dates
  • There can be no date jumps (eg. 2014-11-01 | 2014-11-03)
• Numeric counts
  • Each cell must be a positive integer (eg. no NA or -1)

Reference

An article that talks to the benefit of data generated by overtime.

Notes

This package is in development. Functionality and documentation is being improved and polished.