Crash Course: Machine Learning Pipeline and API using R

Explore rsample in tidymodels to split data into training and testing sets with initial_split, stratified sampling, and cross-validation readiness, preparing ames_train and ames_test for model evaluation.

Learn to preprocess data in R using the tidymodels recipes package, applying normalization, dummy encoding, and other step functions, and bake the recipe on training data to avoid leakage.

Explore feature engineering with recipes and step_functions, using step_other with a 0.05 threshold, step_nzv, and ames_train. Create total_area from area columns, apply step_mutate and step_dummy.

Learn to train and fit models with the parsnip unified API, using rand_forest with ranger or spark for regression, then prep data and predict on ames_test.

Explore hyperparameter tuning for a random forest in R, using vfold_cv cross-validation, the tune package, and grid_random grids with recipes, parsnip, and dials.

Examine tuning results across 10 hold-out sets, compare rmse and rsquared with collect_metrics, and use select_best to identify the best hyperparameters for a parsnip rand_forest model.

Define and tune a regression workflow by combining a preprocessor recipe with a random forest model, tune hyperparameters, and finalize before fitting and predicting.

Leverage workflowsets to compare multiple workflows of preprocessors and models, using recipes like normalize and dummy, with feature engineering such as total_area in R.

Explain how to work with workflow set objects and tuning results, generate entropy-based grids, and identify the best random forest workflow using collect_metrics and RMSE or R-squared.

Build a machine learning pipeline with plumber and tidymodels, set up data and folders, and expose API endpoints for training, predictions, and reporting on a local server.

Learn to define R functions as api endpoints with plumber, install and run plumber, and test a simple echo endpoint via Swagger and curl.

Explore the basics of the pipeline, including rmarkdown dynamic reports driven by params, selective output in code chunks, library checks, and loading the ames housing data with a default summary.

Explore building a machine learning pipeline in R, from seeding data and creating training/test splits to cross-validation, preprocessing with recipes, imputation, and tuning for random forests and linear regression.

Identify the best workflow by rmse using grid_results and rank_results, then generate test predictions with last_fit and plot observed vs predicted with ggplot2, saving the final model to an Rds.

learn to build a plumber api in R to expose a predict_data endpoint using parsnip and read data and the model with read.csv and readRDS to generate predictions.

Explore how the report endpoint in a plumber-based machine learning pipeline renders an html report by rendering an rmarkdown document with input data, target, and tune params.

Interact with a machine learning pipeline via start_api.R and swagger, run the report, tune parameters, and make predictions on unseen data, while inspecting results and feature importance.

Add a new preprocessor with PCA to reduce feature dimensionality using an 80 percent explained variance threshold, integrated into the pipeline.

Add a regression tree to the machine learning pipeline using rpart, tune parameters, and customize the workflow with new models and tune grids; inspect outputs via the API.

Celebrates the completion of a regression pipeline course using tidymodels, workflowsets, and plumber in r. Wishing you the best as you continue your journey as an r programmer.