Entroscope

The story

While building another project, I noticed something interesting. When a food started trending in regional search, the entropy of its signal would quietly drop before it went mainstream. That made it a useful early signal, but every time I wanted to use it, I had to rewrite the same code from scratch. Other entropy measures were even harder, since I had to copy formulas out of research papers and second-guess the math.

So I decided to write it once and do it properly. I built a Python library that puts seven entropy measures behind the same small interface, so switching from one measure to another is a one-word change. It works directly with pandas and NumPy, includes rolling and delta versions of every measure, and has built-in plotting to spot sudden shifts in a signal. I checked the trickier measures against known mathematical results to make sure they were correct. It's now published on PyPI, so anyone can install the tool I wish I'd had from the start.