arff-generator

Read this in Español.

An ARFF (Attribute-Relation File Format — the dataset format used by Weka) file generator that produces randomized data from the attributes and values you define.

Handy for quickly creating test datasets without writing the .arff by hand.

Requirements

• Python 3.6+ (standard library only: random, os)

Usage

Define the relation, the attributes with their possible values, and the classification attribute, then call randomize():

from arffclass import *

file = arffGenerator()
file.setDatas(20)                  # number of data rows to generate
file.setRelation("meteoritos")     # relation name
file.setFileName("prueba")         # output file name (.arff is appended)

# possible values of the classification attribute
file.setClass(["Coronis", "Eos", "Temis", "Flora", "Maria"])

# attribute and its possible values
file.setAttribute("tiporoca", ["arcilla", "niquel", "hierro", "desconocido"])
file.setAttribute("diametro", [1, 2, 3, 4, 5])
file.setAttribute("masa", [2.7, 2.8, 2.9, 3.0, 3.1, 3.2, 3.3])
file.setAttribute("forma", ["dentada", "irregular"])
file.setAttribute("riezgoimpacto", ["bajo", "medio", "alto"])

file.randomize()   # generates the .arff file with random data

Run:

python main.py

Generates output/prueba.arff. The output/ folder is created automatically; if the file already exists, it is overwritten.

Example output

@relation meteoritos

@attribute tiporoca {arcilla, niquel, hierro, desconocido}
@attribute diametro {1, 2, 3, 4, 5}
@attribute masa {2.7, 2.8, 2.9, 3.0, 3.1, 3.2, 3.3}
@attribute forma {dentada, irregular}
@attribute riezgoimpacto {bajo, medio, alto}
@attribute class {Coronis, Eos, Temis, Flora, Maria}

@data
desconocido, '5', '3.1', dentada, alto, Temis
arcilla, '3', '3.1', irregular, bajo, Flora
hierro, '5', '3.2', irregular, medio, Eos
hierro, '2', '2.9', irregular, bajo, Maria
niquel, '3', '2.7', irregular, alto, Eos
niquel, '4', '2.9', irregular, medio, Temis
desconocido, '5', '3.1', dentada, alto, Maria
hierro, '3', '2.9', irregular, bajo, Flora
...

Numeric values are written in single quotes ('3') so Weka treats them as nominal, like the rest of the attributes.

API — arffGenerator

Method	Description
setDatas(int)	Number of data rows to generate
setRelation(str)	Relation name (@relation)
setFileName(str)	Output file name (.arff is appended)
setClass([...])	Possible values of the class classification attribute
setAttribute(name, [...])	Adds an attribute and its list of possible values
randomize()	Generates the .arff file with the random data

Each set* validates the input type and raises an exception if it is wrong (e.g. setDatas requires int, setAttribute requires a list of values).

Files

File	Purpose
arffclass.py	arffGenerator class (reusable API)
main.py	Example using the class
simpleconsole.py	Standalone script version, no class (configured via variables)

Notes

• simpleconsole.py ships with datas = 10000000 by default — would generate a huge file. Adjust that value before running it.
• The classification attribute name is always class; randomize() adds it automatically from setClass().

---

Created by @albertolicea00