0

To create synthetic data there are two approaches:

  1. Drawing values according to some distribution or collection of distributions

  2. Agent-based modelling

For the first approach we can use the numpy.random.choice function which gets a dataframe and creates rows according to the distribution of the data frame.

I wanted to ask if there is a defined function for the second approach "Agent-based modelling" in python or have we implement it on ourself?

Code Pope
  • 5,075
  • 8
  • 26
  • 68
  • 1
    What you ask is unclear. The assertion you made for the approaches is, at best, partial & incomplete (source?); agent-based modelling is a whole scientific topic in itself (see tag description for starters)... – desertnaut Sep 05 '18 at 09:21
  • @desertnaut Everywhere I read about synthetic data for ML these two approaches are mentioned. Here one source: https://www.codementor.io/ericlefort/my-thoughts-on-synthetic-data-kq719a5ss#background-on-synthetic-data – Code Pope Sep 05 '18 at 11:11
  • That's very surprising; *simulation*, in general (of which ABM is a branch) yes, but ABM in particular? Sounds bogus... – desertnaut Sep 05 '18 at 11:13
  • @desertnaut I think he means that we use the given data to train a model (ABM) which after that is able to output synthetic data – Code Pope Sep 05 '18 at 11:24
  • And here another source: https://www.riaktr.com/synthetic-data-become-major-competitive-advantage/ – Code Pope Sep 05 '18 at 11:27
  • ABM models are not trained, and they (normally) do not take data as an input... – desertnaut Sep 05 '18 at 14:07

0 Answers0