Package index
Model fitting
One fitter per model family. psvr_mape() covers the epsilon-SVR models (1 and 2), psvr_rmspe() the LS-SVR models (3 and 4); within each, sym_type = "none" | "even" | "odd" selects the symmetric variant. They are separate functions because the two families share no solver, no dual structure and no hyperparameter search space.
-
psvr_mape() - Fit an epsilon-SVR with MAPE loss
-
psvr_rmspe() - Fit a least-squares SVR with RMSPE loss
-
make_kernel() - Create a kernel function
Cross-validation
Fold-wise fitting for loss = "mape", carrying the converged SMO dual variables from one fold into the next as a warm start.
-
psvr_cv() - Cross-validate psvr_mape() with automatic warm-start across folds
tidymodels / parsnip interface — ε-SVR with MAPE (Models 1 & 2)
Parsnip model specifications for the ε-SVR family (one per kernel type). sym_type = "none" fits Model 1; "even" / "odd" fit Model 2.
-
psvr_mape_rbf()psvr_mape_poly()psvr_mape_linear() - Parsnip model specs: epsilon-SVR with MAPE loss (Model 1)
tidymodels / parsnip interface — LS-SVR with RMSPE (Models 3 & 4)
Parsnip model specifications for the LS-SVR family (one per kernel type). sym_type = "none" fits Model 3; "even" / "odd" fit Model 4.
-
psvr_rmspe_rbf()psvr_rmspe_poly()psvr_rmspe_linear() - Parsnip model specs: LS-SVR with RMSPE loss (Model 3)
-
margin_percentage() - Insensitivity margin in percentage units
-
sigma_heuristic() - Median-distance heuristic for RBF kernel bandwidth
-
rbf_sigma_psvr() - RBF sigma parameter for psvr models
-
rbf_sigma_psvr_data() - RBF sigma parameter with data-driven range for psvr models
-
psvr_option_add() - Apply data-driven rbf_sigma to all psvr workflows in a workflow set
-
psvr_option_add_cost_ls() - Apply data-driven LS-SVR cost range to all m3/m4 workflows in a workflow set
-
cost_psvr() - Cost parameter with extended range for psvr models
-
cost_psvr_ls_data() - Data-driven cost range for LS-SVR psvr models
-
sym_type_param() - Dials parameter for symmetry type
Fit-object methods
Methods for the four fit classes — psvr_mape, psvr_mape_sym, psvr_rmspe, psvr_rmspe_sym. These are what psvr_mape() and psvr_rmspe() return and what the parsnip engine fit wrappers return, so a direct fit and a parsnip fit unwrapped with parsnip::extract_fit_engine() are the same object.
-
predict(<psvr_mape>) - Predict from a fitted epsilon-SVR with MAPE model
-
predict(<psvr_mape_sym>) - Predict from a fitted symmetric epsilon-SVR with MAPE model
-
predict(<psvr_rmspe>) - Predict from a fitted LS-SVR with RMSPE model
-
predict(<psvr_rmspe_sym>) - Predict from a fitted symmetric LS-SVR with RMSPE model
-
print(<psvr_mape>) - Print method for psvr_mape objects
-
print(<psvr_mape_sym>) - Print method for psvr_mape_sym objects
-
print(<psvr_rmspe>) - Print method for psvr_rmspe objects
-
print(<psvr_rmspe_sym>) - Print method for psvr_rmspe_sym objects
-
coef(<psvr_mape>) - Extract coefficients from a psvr_mape model
-
coef(<psvr_mape_sym>) - Extract coefficients from a psvr_mape_sym model
-
coef(<psvr_rmspe>) - Extract coefficients from a psvr_rmspe model
-
coef(<psvr_rmspe_sym>) - Extract coefficients from a psvr_rmspe_sym model
-
summary(<psvr_mape>) - Summarize a fitted epsilon-SVR with MAPE loss
-
summary(<psvr_mape_sym>) - Summarize a fitted symmetric epsilon-SVR with MAPE loss
-
summary(<psvr_rmspe>) - Summarize a fitted LS-SVR with RMSPE loss
-
summary(<psvr_rmspe_sym>) - Summarize a fitted symmetric LS-SVR with RMSPE loss
-
fitted(<psvr_mape>)fitted(<psvr_mape_sym>)fitted(<psvr_rmspe>)fitted(<psvr_rmspe_sym>) - Extract training fitted values from a psvr model
-
residuals(<psvr_mape>)residuals(<psvr_mape_sym>)residuals(<psvr_rmspe>)residuals(<psvr_rmspe_sym>) - Extract training residuals from a psvr model