tests/testthat/test-algo-arima_boost-auto_arima.R

In modeltime: The Tidymodels Extension for Time Series Modeling

# ---- STANDARD ARIMA ----
context("TEST arima_boost: auto_arima_xgboost")


# SETUP ----

# Data
m750 <- timetk::m4_monthly %>% dplyr::filter(id == "M750")

# Split Data 80/20
splits <- rsample::initial_time_split(m750, prop = 0.8)

# Model Spec
model_spec <- arima_boost(
    seasonal_period          = 12,
    non_seasonal_ar          = 3,
    non_seasonal_differences = 1,
    non_seasonal_ma          = 3,
    seasonal_ar              = 2,
    seasonal_differences     = 1,
    seasonal_ma              = 2,
    mtry  = 25,
    trees = 250,
    min_n = 4,
    learn_rate = 0.1,
    tree_depth = 7,
    loss_reduction = 0.4,
    sample_size = 0.9
) %>%
    parsnip::set_engine("auto_arima_xgboost")


# PARSNIP ----

# * NO XREGS ----


# TESTS
test_that("arima_boost: Arima, (No xregs), Test Model Fit Object", {

    skip_on_cran()

    # Fit Spec
    model_fit <- model_spec %>%
        fit(log(value) ~ date, data = rsample::training(splits))

    # Predictions
    predictions_tbl <- model_fit %>%
        modeltime_calibrate(rsample::testing(splits)) %>%
        modeltime_forecast(new_data = rsample::testing(splits))

    # TEST

    expect_s3_class(model_fit$fit, "auto_arima_xgboost_fit_impl")

    # $fit

    expect_s3_class(model_fit$fit$models$model_1, "Arima")

    expect_s3_class(model_fit$fit$data, "tbl_df")

    expect_equal(names(model_fit$fit$data)[1], "date")

    expect_null(model_fit$fit$extras$xreg_recipe)

    # $fit xgboost

    expect_identical(model_fit$fit$models$model_2, NULL)

    # $preproc

    expect_equal(model_fit$preproc$y_var, "value")


    # PREDICTIONS

    # Structure
    expect_identical(nrow(rsample::testing(splits)), nrow(predictions_tbl))
    expect_identical(rsample::testing(splits)$date, predictions_tbl$.index)

    # Out-of-Sample Accuracy Tests

    resid <- rsample::testing(splits)$value - exp(predictions_tbl$.value)

    # - Max Error less than 1500
    expect_lte(max(abs(resid)), 1500)

    # - MAE less than 700
    expect_lte(mean(abs(resid)), 700)

})

# * XREGS ----

# TESTS
test_that("arima_boost: Arima, (XREGS), Test Model Fit Object", {

    skip_on_cran()

    # SETUP

    # Fit Spec
    model_fit <- model_spec %>%
        fit(log(value) ~ date + as.numeric(date) + lubridate::month(date, label = TRUE), data = rsample::training(splits))

    # Predictions
    predictions_tbl <- model_fit %>%
        modeltime_calibrate(rsample::testing(splits)) %>%
        modeltime_forecast(new_data = rsample::testing(splits))

    expect_s3_class(model_fit$fit, "auto_arima_xgboost_fit_impl")

    # Structure

    expect_s3_class(model_fit$fit$data, "tbl_df")

    expect_equal(names(model_fit$fit$data)[1], "date")

    expect_true(!is.null(model_fit$fit$extras$xreg_recipe))

    # $fit arima

    expect_s3_class(model_fit$fit$models$model_1, "Arima")

    # $fit xgboost

    expect_s3_class(model_fit$fit$models$model_2, "xgb.Booster")

    expect_identical(model_fit$fit$models$model_2$params$eta, 0.1)

    expect_identical(model_fit$fit$models$model_2$params$max_depth, 7)

    expect_identical(model_fit$fit$models$model_2$params$gamma, 0.4)

    expect_identical(model_fit$fit$models$model_2$params$colsample_bytree, 1)

    expect_identical(model_fit$fit$models$model_2$params$min_child_weight, 4)

    expect_identical(model_fit$fit$models$model_2$params$subsample, 0.9)

    expect_identical(model_fit$fit$models$model_2$params$objective, "reg:squarederror")

    # $preproc

    expect_equal(model_fit$preproc$y_var, "value")

    # PREDICTIONS

    # Structure
    expect_identical(nrow(rsample::testing(splits)), nrow(predictions_tbl))
    expect_identical(rsample::testing(splits)$date, predictions_tbl$.index)

    # Out-of-Sample Accuracy Tests

    resid <- rsample::testing(splits)$value - exp(predictions_tbl$.value)

    # - Max Error less than 1500
    expect_lte(max(abs(resid)), 1500)

    # - MAE less than 700
    expect_lte(mean(abs(resid)), 700)

})


# ---- WORKFLOWS ----


# TESTS
test_that("arima_boost: Arima (workflow), Test Model Fit Object", {

    skip_on_cran()

    # SETUP
    # Model Spec
    model_spec <- arima_boost(
        seasonal_period          = 12,
        non_seasonal_ar          = 3,
        non_seasonal_differences = 1,
        non_seasonal_ma          = 3,
        seasonal_ar              = 1,
        seasonal_differences     = 1,
        seasonal_ma              = 1,
        mtry  = 25,
        trees = 250,
        min_n = 4,
        learn_rate = 0.1,
        tree_depth = 7,
        loss_reduction = 0.4,
        sample_size = 0.9
    ) %>%
        parsnip::set_engine("auto_arima_xgboost")

    # Recipe spec
    recipe_spec <- recipes::recipe(value ~ date, data = rsample::training(splits)) %>%
        recipes::step_log(value, skip = FALSE) %>%
        recipes::step_date(date, features = "month") %>%
        recipes::step_mutate(date_num = as.numeric(date))

    # Workflow
    wflw <- workflows::workflow() %>%
        workflows::add_recipe(recipe_spec) %>%
        workflows::add_model(model_spec)

    wflw_fit <- wflw %>%
        fit(rsample::training(splits))

    # Forecast
    predictions_tbl <- wflw_fit %>%
        modeltime_calibrate(rsample::testing(splits)) %>%
        modeltime_forecast(new_data = rsample::testing(splits), actual_data = rsample::training(splits)) %>%
        dplyr::mutate(dplyr::across(.value, exp))


    # TESTS

    expect_s3_class(wflw_fit$fit$fit$fit, "auto_arima_xgboost_fit_impl")

    # Structure

    expect_s3_class(wflw_fit$fit$fit$fit$data, "tbl_df")

    expect_equal(names(wflw_fit$fit$fit$fit$data)[1], "date")

    expect_true(!is.null(wflw_fit$fit$fit$fit$extras$xreg_recipe))

    # $fit arima

    expect_s3_class(wflw_fit$fit$fit$fit$models$model_1, "Arima")

    # $fit xgboost

    expect_s3_class(wflw_fit$fit$fit$fit$models$model_2, "xgb.Booster")

    expect_identical(wflw_fit$fit$fit$fit$models$model_2$params$eta, 0.1)

    expect_identical(wflw_fit$fit$fit$fit$models$model_2$params$max_depth, 7)

    expect_identical(wflw_fit$fit$fit$fit$models$model_2$params$gamma, 0.4)

    expect_identical(wflw_fit$fit$fit$fit$models$model_2$params$colsample_bytree, 1)

    expect_identical(wflw_fit$fit$fit$fit$models$model_2$params$min_child_weight, 4)

    expect_identical(wflw_fit$fit$fit$fit$models$model_2$params$subsample, 0.9)

    expect_identical(wflw_fit$fit$fit$fit$models$model_2$params$objective, "reg:squarederror")


    # $preproc
    mld <- wflw_fit %>% workflows::extract_mold()
    expect_equal(names(mld$outcomes), "value")

    # PREDICTIONS

    full_data <- dplyr::bind_rows(rsample::training(splits), rsample::testing(splits))

    # Structure
    expect_identical(nrow(full_data), nrow(predictions_tbl))
    expect_identical(full_data$date, predictions_tbl$.index)

    # Out-of-Sample Accuracy Tests
    predictions_tbl <- predictions_tbl %>% dplyr::filter(.key == "prediction")
    resid <- rsample::testing(splits)$value - predictions_tbl$.value

    # - Max Error less than 1500
    expect_lte(max(abs(resid)), 1500)

    # - MAE less than 700
    expect_lte(mean(abs(resid)), 700)

})