The tidysynthesis R package
summary
Built an R package called tidysynthesis that uses parsnip and recipes from tidymodels to fit a sequential ML model for each variable, sampling predictions from terminal nodes via sample_rpart rather than returning conditional means. Produced two synthetic IRS files (a 20-variable supplemental SynPUF and a >200-variable SynPUF with hundreds of thousands of records and 25 strata). Adds extra noise, enforces logical constraints, supports replicates and parallel runs, and reproduces univariate and bivariate patterns closely.
AI-written summary · about
chapters (4)
- 0:00Start
- 4:25Supplemental SynPUF
- 8:53What Algorithms Should We Use?
- 12:24The Synthetic Data are Similar to the Confidential Data
related talks
18mtidymodels: Adventures in Rewriting a Modeling Pipeline
23mparsnip A tidy model interface
57mGood Practices for Applied Machine Learning
23mTotal Tidy Tuning Techniques
21mTidypredict with recipes, turn workflow to SQL, spark, duckdb and beyond
19mtidymodels for time-to-event data
31mUsing Simpson’s Paradox to Discover Interesting Patterns in...
18mCensored - Survival Analysis in Tidymodels