Distribution analysis

Ask the Community
1

I am trying to identify the distribution of different subsets of data. Different materials/items have different behaviour in waste across restaurants. I am trying to identify what form of distribution those different items follow. Are they normally distributed, left skewed, poisson, weibull, binomial …
Has someone done this in Foundry before? Any pointers, ideas? Ideally i can do this in Builder? Anyone used some ML models to try and help w this?

4

Hey @emuellermohl I’ll let other people answer on the ML models but something basic you could do in builder is utilize the mean, mode, standard deviation boards aggregations and then use those to calculate your distributions.

For example, using these definitions:

  • Normal Distribution: The mean and mode are approximately equal, and the data is symmetrically distributed around the mean.
  • Left Skew (Negative Skew): The mode is greater than the mean, indicating a longer left tail.
  • Right Skew (Positive Skew): The mean is greater than the mode, indicating a longer right tail.

code:

if abs(mean_value - mode_value) < 0.5 * stddev_value:
    distribution_type = "Normal Distribution"
elif mean_value < mode_value:
    distribution_type = "Left Skew (Negative Skew)"
elif mean_value > mode_value:
    distribution_type = "Right Skew (Positive Skew)"
else:
    distribution_type = "Unknown Distribution"

This might be more basic than what you’re looking for but could be a start depending on what distributions you care about