This platform is an interactive research compedium of my academic publication below.
Gürol Canbek (2022). G-Category: A novel method to quantifying and categorizing data sets. Journal of Machine Learning Research (To be submitted).
The platform provides ready-to-run open-source R scripts for the new method called G-Category (Greatness Category). The method is proposed in the article above to categorize the sizes of a group of data sets in two dimensions: sample space and feature space. The G-Categories are small, medium, shallow, skinny, and large. An experimenter is prepared to test the G-Category method in example synthetic (linear and random size distributions) and the real data sets found in the literature.
The results are given for two approaches: pure geometric (correct) approach and pure arithmetic (erroneous) approach to see the difference. Refer to the article for more information.
Note: Please, cite my article if you would like to use and/or adapt the code, datasets, methodology, and other materials provided and let us know. Thank you for your interest.
Skip to Quick Start section below to learn how to use this platform.
You can calculate the G-Categories of your own group of data sets in R using our scripts. Just do the following six steps:
- First, copy our two R scripts (gcategory.R and powerstat.R) in your folder.
- Second, include our main script (i.e.
source('gcategory.R')) in your script file or in R interactive console - Third, store the sample sizes of your data sets in a vector (e.g.
n <- c(100, 200, 300)) - Fourth, store the corresponding feature space sizes of your data sets in another vector (e.g.
m <- c(10, 12, 13)) - Fifth, name the corresponding data sets (e.g.
DSs <- c('DS1', 'DS2', 'DS3')) - Finally, use the provided functions (such as
greatnessCategories,plotTableGCsDetailed,plotGraphGCs)
A minimal example:
# Put the gcategory.R and powerstat.R script fiiles in your current directory
source('gcategory.R')
# Sample space sizes
n <- c(100, 200, 300)
# Feature space sizes
m <- c(10, 12, 13)
# Data set names
DSs <- c('DS1', 'DS2', 'DS3')
# Using default (correct) approach (pure geometric) (power=0 and theta=1)
greatnessCategories(m, n)
tabulateGCs(m, n, DSs)The outputs are
[1] "Small" "Medium" "Large"
10 12 13
300 DS3 (Large)
200 DS2 (Medium)
100 DS1 (Small)
This platform provides Data on the bottom-left, Code on the top-left, and Results on the right pane.
You can explore any file by clicking. The results of a pre-runned experimentation elaborated in the article is already provided in Results pane.
If you would like to experiment on your own, you can;
- Click Run botton
on the right of the top toolbar to launch experimentation. After the run is finished, the files (tabular data and graphics) are populated in the Results pane at the right for your review.
The original code repository and future updates can be found at https://github.com/gurol/gcategory
├── code
│ ├── Experimenter.R : Experiment G-Category method in synthetic and real data sets
│ │ (total five data sets).
│ ├── gcategory.R : The module for calculating G-Categories
│ ├── LICENSE : License file
│ ├── main.R : Starter R script (internal file for this platform)
│ ├── powerstat.R : Script for calculating several statistics such as mean, standard
│ │ deviation, z-scored based on the power coefficient.
│ ├── README.md : This help file
│ └── run.sh : Shell script (internal file for this platform)
│
├── data
│ └── (No Data)
│
└── results
├── output : Output log of the experimentation (showing the steps)
├── 1_SyntheticDSs_Linear : The folder holding the results for the synthetic
│ data sets having linear space size distributions.
├── 2_SyntheticDSs_Random : The folder holding the results for the synthetic
│ data sets having random space size distributions.
├── 3_BenignDSs : The folder holding the results for the real data sets in the
│ literature (Android benign application samples).
├── 4_MalignDSs : The folder holding the results for the real data sets in the
│ literature (Android malign application (malware) samples).
├── 5_MalwareFamilyDSs : The folder holding the results for the real data sets in
│ the literature (Android malign application (malware) samples
│ having malware family information or the recent samples).
│
└── [in each folder above ("n" is the configuration number)]
├── n(ari/geo)_a(DataSetsName).png : G-Categories calculated via arithmetic/
│ geometric approach. It shows detailed
│ information per data set such as Z-scores.
├── n(ari/geo)_b(DataSetsName)Graph.png : G-Categories calculated via arithmetic/
│ geometric approach are shown in space graph
├── n(ari/geo)_c(DataSetsName)Combination.png : G-Categories calculated for the
│ data sets having all the combination of the
│ space sizes (via arithmetic/
│ geometric approach)
├── n(ari/geo)_d(DataSetsName).csv : Tabulated G-Categories calculated via
│ arithmetic/geometric approach
└── n(ari/geo)_e(DataSetsName)Combination.csv : Tabulated G-Categories calculated
for the data sets having all the combination
of the space sizes (via arithmetic/geometric
approach)
Copyright (C) 2017-2022 Gürol Canbek
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent