Function reference

Modeltime Workflow

The main workflow functions for time series modeling.

Core Functions

modeltime_table() as_modeltime_table(): Scale forecast analysis with a Modeltime Table

modeltime_calibrate(): Preparation for forecasting

modeltime_forecast(): Forecast future data

modeltime_accuracy(): Calculate Accuracy Metrics

modeltime_refit(): Refit one or more trained models to new data

modeltime_fit_workflowset(): Fit a workflowset object to one or multiple time series

Recursive Forecast Prediction

recursive(): Create a Recursive Time Series Model from a Parsnip or Workflow Regression Model

panel_tail(): Filter the last N rows (Tail) for multiple time series

Plotting & Tables

plot_modeltime_forecast(): Interactive Forecast Visualization

plot_modeltime_residuals(): Interactive Residuals Visualization

table_modeltime_accuracy(): Interactive Accuracy Tables

Residual Analysis

modeltime_residuals(): Extract Residuals Information

modeltime_residuals_test(): Apply Statistical Tests to Residuals

plot_modeltime_residuals(): Interactive Residuals Visualization

Nested Forecasting

Forecast many time series iteratively using “nested modeltime tables”. Used to apply models to each time series panel independently.

Core functions

modeltime_nested_fit(): Fit Tidymodels Workflows to Nested Time Series

modeltime_nested_select_best(): Select the Best Models from Nested Modeltime Table

modeltime_nested_refit(): Refits a Nested Modeltime Table

modeltime_nested_forecast(): Modeltime Nested Forecast

Extractors

extract_nested_test_accuracy() extract_nested_test_forecast() extract_nested_error_report() extract_nested_best_model_report() extract_nested_future_forecast() extract_nested_modeltime_table() extract_nested_train_split() extract_nested_test_split(): Log Extractor Functions for Modeltime Nested Tables

Workflow

extend_timeseries() nest_timeseries() split_nested_timeseries(): Prepared Nested Modeltime Data

Algorithms

The parsnip-adjacent algorithms that implement time series models.

Core Forecasting Methods

These models come with modeltime.

prophet_reg(): General Interface for PROPHET Time Series Models

prophet_boost(): General Interface for Boosted PROPHET Time Series Models

arima_reg(): General Interface for ARIMA Regression Models

arima_boost(): General Interface for "Boosted" ARIMA Regression Models

exp_smoothing(): General Interface for Exponential Smoothing State Space Models

seasonal_reg(): General Interface for Multiple Seasonality Regression Models (TBATS, STLM)

nnetar_reg(): General Interface for NNETAR Regression Models

Additional Algorithms

These algorithms have additional dependencies that can be installed with dependencies = TRUE

adam_reg(): General Interface for ADAM Regression Models

temporal_hierarchy(): General Interface for Temporal Hierarchical Forecasting (THIEF) Models

Baseline Algorithms (Simple Methods)

window_reg(): General Interface for Window Forecast Models

naive_reg(): General Interface for NAIVE Forecast Models

Parallel Processing

parallel_start() parallel_stop(): Start parallel clusters using parallel package

control_refit() control_fit_workflowset() control_nested_fit() control_nested_refit() control_nested_forecast(): Control aspects of the training process

create_model_grid(): Helper to make parsnip model specs from a dials parameter grid

Modeltime Workflow Helpers

combine_modeltime_tables(): Combine multiple Modeltime Tables into a single Modeltime Table

add_modeltime_model(): Add a Model into a Modeltime Table

drop_modeltime_model(): Drop a Model from a Modeltime Table

update_modeltime_model(): Update the model by model id in a Modeltime Table

update_model_description() update_modeltime_description(): Update the model description by model id in a Modeltime Table

pluck_modeltime_model() pull_modeltime_model(): Extract model by model id in a Modeltime Table

pull_modeltime_residuals(): Extracts modeltime residuals data from a Modeltime Model

pull_parsnip_preprocessor(): Pulls the Formula from a Fitted Parsnip Model Object

Accuracy Metrics (Yardstick)

Metric Sets and Summarizers

default_forecast_accuracy_metric_set() extended_forecast_accuracy_metric_set(): Forecast Accuracy Metrics Sets

summarize_accuracy_metrics(): Summarize Accuracy Metrics

New Accuracy Metrics

maape(): Mean Arctangent Absolute Percentage Error

maape_vec(): Mean Arctangent Absolute Percentage Error

Parameters (Dials)

The dials parameter functions that support hyperparameter tuning with tune.

General Time Series

seasonal_period(): Tuning Parameters for Time Series (ts-class) Models

NNETAR

num_networks(): Tuning Parameters for NNETAR Models

ADAM

use_constant() regressors_treatment() outliers_treatment() probability_model() distribution() information_criteria() select_order(): Tuning Parameters for ADAM Models

Temporal Hierachical Models

combination_method() use_model(): Tuning Parameters for TEMPORAL HIERARCHICAL Models

Developer Tools

Tools for extending modeltime.

new_modeltime_bridge(): Constructor for creating modeltime models

create_xreg_recipe(): Developer Tools for preparing XREGS (Regressors)

juice_xreg_recipe() bake_xreg_recipe(): Developer Tools for processing XREGS (Regressors)

parse_index_from_data() parse_period_from_index(): Developer Tools for parsing date and date-time information

get_model_description(): Get model descriptions for parsnip, workflows & modeltime objects

get_arima_description(): Get model descriptions for Arima objects

get_tbats_description(): Get model descriptions for TBATS objects

Data

m750: The 750th Monthly Time Series used in the M4 Competition

m750_models: Three (3) Models trained on the M750 Data (Training Set)

m750_splits: The results of train/test splitting the M750 Data

m750_training_resamples: The Time Series Cross Validation Resamples the M750 Data (Training Set)