ML Models

Contains code to build, train, package and deploy a model.

ML Pipelines Overview:

Data (reproducible -> R) --> Data Analysis --> Data Pre-processing(R) (Feature Engineering) --> Variable Selection(R)(Feature Selection) --> ML model building(R) --> Model deploy(R)

Data Layer(base) --> Feature layer --> Scoring layer --> Evaluation layer

Models in this repo:

Titanic

• Classifier model based on Logistic regression to predict survivors in Titanic
• Kaggle link - https://www.kaggle.com/datasets/heptapod/titanic

Customer Behavior

• Given customer details, this tensorflow-keras/pytorch neural network (with 2 hidden layers) based model predicts the probability of the customer likeliness to buy or not

Restaurant reviews

• Given a review, this NLP model based on K-NN/tensorflow-keras/pytorch neural network predicts whether it is positive review or not.
• Kaggel link - https://www.kaggle.com/datasets/d4rklucif3r/restaurant-reviews
• (Tests developed using pytest framework)

Others in this repo:

Tensorflow JS

• Develop ML models in JavaScript, and use ML directly in the browser or in Node.js
• This repo has a html file which uses tfjs to run a model in the client browser
• https://www.tensorflow.org/js

MLFlow

• This repo has KNN(sklearn) and Pytorch models integrated with mlflow.
• Opensource app for ML lifecycle management
• MLflow is an open source platform to manage the ML lifecycle, including experimentation, reproducibility, deployment, and a central model registry.
• https://mlflow.org/

References:

Models are developed and structured based on learnings from:

• https://www.udemy.com/course/deployment-of-machine-learning-models/
• https://www.udemy.com/course/machine-learning-deep-learning-model-deployment/