Randomizer

Randomizer

This application plots a graph that mimics the real curve. A pythonscript that states 4 times above 80% and the quarters being 7 quarters apart. The visual plot is handy to check the performance of the script. As I found out later, it is much easier to let an LLM describe the differences, based on the distinction have it generate a prompt to have it in another go, generate a python script.

labelling

While using synthetic data, data that fits a certain pattern, we can attach a label to this data. Each plot generated this way can get the label “matches pattern”

Training data

We can generate as much data as we want, both data that fits the the pattern, and completely random data (label no match) This method allows to train a model (keras)

Filter

OK, we have a model, and we can verify our real data (starting with Deutz, which should match the pattern)