AutoML Pipeline

Automatic Machine Learning Pipeline

• https://seththedatascientist-automl.streamlit.app/

• I created a streamlit application to automate a machine learning pipeline for analyzing and modeling from a dataset. Key Features include:

  • Data ingestion (handles multiple file types)

  • Exploratory data analysis (handles missing values, feature and target distributions, correlation matrix, etc)

  • Feature extraction and transformation (handles scaling, imputing, and can create PCA components)

  • Automatic detection of regression or classification problems

  • Training multiple models with hyperparameter tuning and cross validation steps

  • Selecting and saving best model for making predictions on test set (will also generate predictions from test set if provided)

  • Generates HTML report for modeling process as well EDA output that can be downloaded as zip file

• Future work will include incorporating more model types including neural networks (with automatic detection of most effective model architecture), time series predictions, clustering algorithms, and ensemble models.