Data Preprocessing & Deployment

[zubin131]

Hi Daniel,

Are there any defined preprocessing steps that can be applied when that data has 100's of features? It's not feasible to identify each column by the name/index and transform it back to numbers. Maybe building some kind of function might help?

Also, can you provide a reference/course for deploying the built model using Tensorflow lite or Tensorflow.js?

Thank you!!

[mrdbourke]

Hey Zubin,

You'll want to look into something like Scikit-Learn's Column Transformer - https://scikit-learn.org/stable/auto_examples/compose/plot_column_transformer_mixed_types.html

Or on the TensorFlow side of things, TensorFlow Transform should help you out - https://www.tensorflow.org/tfx/tutorials/transform/census

As for deployment, it depends where you want to deploy your model.

I made a video on deploying a TensorFlow model to Google Cloud, you may find that helpful - https://youtu.be/fw6NMQrYc6w

Otherwise, I'd also look into Full Stack Deep Learning's materials on deployment - https://youtu.be/quHUi1ENImI

[zubin131]

Awesome, thank you so much Daniel for your prompt response!

It would really be helpful if you could create a course dedicated to just ML and DL projects for real-world business scenarios. It would benefit so many students to kick off their careers and build strong portfolios. Just a thought 👍 :)