- Our ArianeLogiX Concept is based on a domain logic, which we call ArianeLogiX Thread. How do things work in your domain? Describe it along a path and build your ArianeLogiX Thread. It should be simple enough and follow first order principles without getting lost in unnecessary details. As often, the right balance between completeness and efficiency. Dependent on the challenge, a back of the envelope calculation might be enough…
- Of course we know that the problem is more complicated in reality, however, we focus on the parameters which matters the most in order to come quickly to a decision.
- Now, in a world without uncertainty there would be very little business opportunities. High uncertainty often correlates with high risk but also with high reward. It is up to you to understand and quantify ranges of possible outcomes for an informed decision. In order to do that, describe the input values for your ArianeLogiX Thread with uncertainty distributions. We call them Dancing Distributions because they are elegant and highly interactive.
- You are not an expert in statistics? Don’t worry, we make things easy for you with our maths-free distribution builder. No need to choose any pre-set distributions with unknown parameters… Building the Dancing Distributions is like responding to an interview! What is your best guess for a given input value? How good it be? How bad? How sure are you? Interactively create your customized uncertainty distribution!
Learn more about our unique Dancing Distributions.
- In the so-called “Monte Carlo” approach, a calculation (the “Ariane’s Thread”) is repeated hundreds or thousands of times with different sets of input parameters extracted from the uncertainty distribution (our Dancing Distributions). Uncertainty present in any input parameter will propagate through the ArianeLogiX Thread to the results, resulting in an uncertainty distribution of results which describe the ranges of possible outcomes. Those can be used for Probabilistic Predictions, i.e. the extraction of critical values (e.g., the average result, pessimistic result which should be exceeded with a 90% probability, etc.).
- More complex ArianeLogiX Threads can be built, where individual iterations (calculations with one set of parameters extracted from the uncertainty distributions) can take a different path according to rules and thresholds. An ArianeLogiX Thread can have multiple tails, each representing a very different category of outcomes. Every tail has a probability of occurring, which can be translated into a risk parameter.
For a comprehensive non-statistical example of a simple ArianeLogiX Thread, Dancing Distributions and Probabilistic Predictions, see our Introduction to uncertainty and probabilistic modelling.
- Finally, predictions need to meet reality where it is known. Matching results based on forward calculation along the ArianeLogiX Thread (data calibration) can be challenging, in particular in a probabilistic approach. This is where Artificial Intelligence enables structured search for matching parameters combination.
- You are concerned that the machine will take over? No, it will be your best friend. Starting with the set of Dancing Distributions you shaped, it will tell you which parameters really matter, and which do not, and suggest a reduced parameter range within your specification, or, if you allow it, in which direction to step out for a better data match. We believe that to be accepted, Artificial Intelligence must be at your service and help you getting insight.