lethal-toothpaste

Where did the name come from?

I used a random phrase generator online. This one might have actually been directly from github repo name generator, I don't remember

What is this repo?

I got sick of looking for teaching datasets:

• Datasets that were modest in size (100-1000) rows
• More realistic data (real countries, real products, real money values)
• That had some interesting patterns built into them

And while kaggle.com had a number of interesting datasets, (some of which serve as some base information):

• many of them were too confusing to students
• or lacking some basic information to make it easily readable

What techniques are used

Logical Explainable Patterns: Instead of uniform random everything, datasets are "logically generated" for example:

• Generate an item catalog with the following:
  • proposity to view (p = 0 - 1.0)
  • proposity to buy (p = 0 - 1.0)
• generate a list of random users
• give each user a top 2-5 shopping categories, with a uneven distribution (say 0.5, 0.3, 0.2)
• Use that probability profile to select some items from a catalogy under that category

Realistic Profiles:

• For example, randomly selecting a country is based on the GDP distribution
• Selecting a random quantity order follows an inverse relationship:

  1 items = 1 / 1 = 1.0
  2 items = 1 / 2 = 0.5
  5 items = 1 / 5 = 0.2
  ...
  50 items = 1 / 50 = 0.02
  Then normalize all of them so total == 1
  

Trend Control

• Once a group of transactions are created, we can apply sin and cos curves against the data to ensure dips in the middle of the year, or up-trends near the end.