MODEL DEPLOYMENT IN GCP

“No machine learning model is valuable, unless it’s deployed to production.” – Luigi Patruno 

 The machine learning models are being used to achieve many difficult problems. 

For Neuromarketing we have following datasets and for those we have different types of models to be deployed on GCP. 

Dataset Name  Learn from Leaders of IT “How to use Chatgpt and Generative AI” Join the Event Register Now	Model type
FMRI	Deeplearning
Facial Recognition	Deeplearning
Emotion Data	knn
Student MOOC	Random forest
eeg-dataset/eye-tracking	SVM/random forest

The following table has information about the dataset and the type of model deployed on GCP 

 

 

In order To test and evaluate the prebuilt model in GCP for predictions we must follow the following steps: 

1. Upload the model in storage bucket in GCP Make sure to have .joblib file in case of scikit learn model and .pd or .pbtxt kind model file for tensorflow model 
2. Create a GCP model AI model Resource. With the model.joblib file inplace write the following code for data file for model 

Emotion Data

import joblib  joblib.dump(neigh,’model.joblib’)

 Above is the code for emotion dataset with the model name as neigh 

1. Also create AI Platform prediction for version and specify the path where the model is saved.Later to the creation of model in notebook a model must be created in AI resource platform also create a version for the model with all the newest and latest versions for python, framework, and ML runtime.
2. The framework for Student MOOC, eeg and eye-tracking we used Scikit learn, for facial recognition, FMRI and emotion we used XGboost
3. Make sure to give the path for the model.joblib file from bucket and save the details for successful model deployment.
4. After the model is deployed it can be tested and evaluated for future predictions.