0

I have a dataset that looks like this:

product_name brand price_change_percentage release_year
Toy X X 20 2001
Toy Y Y -12 2020

I would like to use Tensorflow to predict price_change_percentage based on product_name, brand, and release_year.

Which kind of Tensorflow model should I conduct and how can I change the non-numerical product_name and brand values into numerical values? Thank you very much!

Shawn Hunter
  • 11
  • 2
  • Try tensorflow_decision_forests – mdaoust Dec 16 '22 at 01:17
  • Thank you! Can I pls ask why? – Shawn Hunter Dec 16 '22 at 01:52
  • @mdaoust, thank you for your suggestion. But I still need to change non-numerical values to numerical values to work with tensorflow_decision_forests... Can you help me suggest the best way to do this? Should I got with One-hot encoding or Word Embedding? Thank you!! – Shawn Hunter Dec 16 '22 at 03:01
  • "But I still need to change non-numerical values to numerical values". No, you don't. Decision forests handle this automatically. Try the google sheets extension: https://www.tensorflow.org/decision_forests/simple_ml_for_sheets – mdaoust Dec 16 '22 at 13:52
  • you will need a RNN/Transformer to convert the name in a fixed size vector, same as for brand (but it's arguable that the name can help your model to predice the price change), and then the usual FFNN.... but for tabular data, probably RF or gradient boosting techniques work better – Alberto Sinigaglia Dec 16 '22 at 14:29

0 Answers0