PortfolioTesteR 0.1.2

New Features

• Tabular ML pipeline
  • roll_fit_predict() — rolling fit/predict for pooled/per-symbol/per-group workflows
  • ml_backtest() — one-call wrapper from features → scores → weights → backtest
  • tune_ml_backtest() — quick hyperparameter sweeps (e.g., top_k, temperature)
• Sequence pipeline (deep learning ready)
  • roll_fit_predict_seq() — rolling fit/predict over flattened sequence features
  • ml_backtest_seq() — sequence wrapper with in-sample normalization
  • cv_tune_seq() — purged/embargoed K-fold CV inside IS window
  • wf_sweep_tabular() — walk-forward sweeps with OOS metric distributions
• Diagnostics
  • membership_stability() — portfolio membership drift analysis
  • turnover_by_date() — turnover time series
  • evaluate_scores(), ic_series(), coverage_by_date() — score quality utilities
  • select_top_k_scores_by_group() — group-aware selection
• Utilities
  • demo_sector_map() and validate_group_map() for group-aware modeling
  • RAM safety guards in sequence builders (max_samples, mem_budget_gb)

Improvements

• Consistent in-sample normalization for ML workflows
• Examples use \donttest{} and @examplesIf() for network/DB access
• Vignettes build via rmarkdown::html_vignette
• Extended globals.R to silence data.table NSE notes
• Documentation improvements and parameter alignment

Compatibility

• No breaking changes — existing strategies continue to work
• Minor API standardization in validate_no_leakage()

PortfolioTesteR 0.1.1

Previous CRAN release

New Features

• Walk-forward optimization: run_walk_forward() with wf_report()
• Parameter grid optimization: run_param_grid()
• S3 print methods for optimization results

Improvements

• Examples wrapped in \donttest{} for CRAN compliance
• SQL examples guard on RSQLite availability
• Added list_examples() and run_example() helpers