Machine Learning Portal

• Goal
• Application
• Workflow
• Process Design
• Live Project

Goal

A complete end-to-end machine learning portal that covers processes starting from model training to model predicting results using FastAPI.

Application

Main Menu

• Dataset: Set default dataset from the list that will be used for training and building the models.
• Train: Start the training process
• Prediction: Upload test dataset to get predictions.

Save sample test data (sample_test.csv for testing the prediction module.

Logger Menu

• Data Preprocessing: Logs generated during preprocessing process.
• Training: Logs generated during training process.
• Prediction: Logs generated during prediction process.

Workflow

• Import/upload dataset
• Preprocessing
  • Categorical features cleaning
  • Handling missing value (categorical)
  • Handling missing value (numeric)
  • Encoding cateogrical feature
  • Over sampling
  • Clustering
  • Train-Test split
  • Model selection
    • Hyper parameter tuning
  • Saving best model
  • Download prediction results

Process Design

Live Project

https://machine-learning-portal.herokuapp.com