Course 4 Week 1 happy_model.evaluate results

Hi,

I ran the happy_model.fit a couple of times and used happy_model.evaluate after each new fit. The results seem to vary quite a lot between 65 - 92 accuracy. Why is that? If that’s on purpose what would I conclude from this in a real life example and what would be my next steps to make evaluate-results more stable?

Thank you!

Best,
Bakir

Hi Bakir,

The happy_model is a simple model which is just used here for educational purposes to show learners how to build and understand models. If you take the rest of the course you will encounter more advanced models. Here is an interesting overview of more elaborate models. Happy learning!