Data Generator

Use this nodejs program to generate random data according to distribution (uniform | normal), sequence, categorical or join.

Scope

You may find open dataset to use for Machine Learning test such us this one (https://www.kaggle.com/datasets/shivamb/machine-predictive-maintenance-classification) but usually they are only training dataset, this force to use the same dataset as training and apply. This program may help you to generate a similar dataset base on the common distribution

Prerequisites

You need nodejs ver. 14.x or newer and npm ver. 8.x or newer (older version may work, but not tested)

Installation

Download the zip file or clone di repo, then from the project folder in a console type

npm install

How to use it

You have to edit the Descriptor.json file, there are 3 first level section which are:

• Fields
• Output
• Format

Once you set up the descriptor.json, from the command line just type

        node index.js <record to produce>

Descriptor.json - configuration

Output

Output contain the array of the output field to print. You may want to generate more fields and join some of them, but then print only a subset (e.g. print the joined result but not the single filed used in the join)

Format

Here you can specify the output format

• CSV
• JSON
• ARRAY

Fields

Fields contains the information on the output to generate. You need to add a section for each field you want to generate, the name of the section in the actual field name. The section need to contain the mandatory property "type" which can be:

• counter
• categorical
• normal
• uniform
• join

Counter

If type is counter, you also need to specify the property start. Counter produce a progressive number starting from the property "start"

Nategorical

If type is categorical, you need to specify other 2 properties: "values" and "likelihood". Categorical produce a list of string in output. The property Values is an array of the string to produce in output, the property likelihood is an array (same dimension of values) which specify the probability for each values: E.g.

        "Class" : {
            "type" : "categorical",
            "values" : ["L","M","H"],
            "likelihood" : [20,30,100]
        }
   

Produce for the field called Class the output L or H ot M Print out L with the probability of 20% Print out M with the probability of 10% (30 - 20) Print out H with the probability of 70% (70 - 30)

Normal

If type is normal you need to specify the decimal position the mean and the variance. Normal print out a random number with the normal (or Gaussian) distribution.

E.g.

        "Air temperature [K]" : {
            "type" : "normal",
            "decimal" : 1,
            "distribution" : {
                "mean" : 300,
                "variance" :4
            }
        }

Produce for the field called "Air temperature [K]" a random number with 1 decimal with the mean of 300 and a variance (standard deviation) of 4.

Uniform

If type is uniform you need to specify the decimal position the from and the to properties. Uniform print out a random number with the uniform distribution (all the number have the same probability)

E.g.

        "Tool wear [min]" :{
            "type" : "uniform",
            "decimal" : 0,
            "distribution" : {
                "from" : 0,
                "to" :250
                
            }

Produce for the field called "Tool wear [min]" a random number between 0 and 250. All the number with the same probability.

Join

If type is join you need to specify "fields" as property. Fields is an array of the fields to join.

E.g.

        "Product ID" : {
            "type" : "join",
            "fields" : ["Type","Prod id num"]
        }

Produce a the field called "Product ID" as concatenation of "Type" and "Prod id num"