SELECTING COMBINATIONS

The interesting thing about variable dimension bridge analysis is that it is a problem that has multiple solutions. Multidimensional datasets equals multiple solutions which, on the bright side πŸ‘, equals multiple ways of quickly looking into the business and hopefully getting useful insights πŸ’‘.

The only way to avoid having multiple solutions is not to use multidimensional datasets in your bridge analysis, which is a pity because you loose a lot of insights. If you do the work in Excel, you won't avoid the problem of multiple solutions, but you might wisely ignore it, given the time it takes to do an iteration πŸ™‚.

With 'Mparanza - not necessarily with Excel πŸ˜‰ - all these solutions will be mathematically correct. Hopefully, a few will be insightful and actionable. How do you find these precious gems πŸ’Ž?

Short answer - not with machine learning. By leveraging on human-machineπŸ‘¨β€πŸ’ΌπŸ€– collaboration. Machine learning and AI are supercool. However as the term implies, the machine needs to "learn".

In order for the computer to learn to recognize a cat 🐱, (i) you need to know what a cat looks like and (ii) you need to show the computer lots of cat pictures πŸ˜ΌπŸ˜ΉπŸ™€πŸ˜ΎπŸ˜ΏπŸ˜»πŸ˜ΊπŸ˜ΈπŸ˜½, and tell it these are all cats, so if it finds something similar, it can tag it as a cat 🐈.

In bridge analysis, the cat we are trying to recognize is an insightful report. It is difficult to prepare a training dataset of insightful reports so the machine can learn to recognize them. Insightful is in the eye of the beholder. We might get there one day. For the moment, 'Mparanza's approach is to make it easy and quick to iteratively tweak the report until we finally get what we want.

We simply look at the file and try to automatically adjust the parameters. Hopefully we get something insightful in the first run. If not, you will need to change the parameters manually and iterate.