Using #MachineLearning to predict the age of a #VIN
TLDR; Source code for this is available on github here; https://github.com/infiniteloopltd/VinAge
This is a first working example of something useful that I’ve created with Microsoft’s ML.NET framework.
I’ve used the “Issue Classification” scenario, training the model on 10,000 known VIN’s and known ages, and then tested it against a VIN that wasn’t in the training data, and it successfully got the age.
It didn’t work so well on a US VIN, but I guess there needs to be more variation in the training data. Also, it took about 1 second per 100 rows to train, so I can imagine some overnight training sessions!
Antipatterns:
I had also tried this on a “Price prediction” scenario, but it gave quite innacurate results, even though the training only took 30 mins over 6 million rows.