Tuning Dimension Smoothers First Version

[AlbertoAlmuinha]

Hi @mdancho84

In this PR I have added the arguments that we discussed in Slack (up to 5) to be able to tune the dimensions of the smoothers. As you can see in this example, each term has the dimension passed as an argument (in order). I have also created the parameters in the file "dials-gen_additive_mod_params.R" to be able to tune.

I haven't had time to test yet but everything should work for tuning (I think). So you can play around a bit and see what you think.

library(tidymodels)
library(modelgam)
library(tidyverse)
library(timetk)
library(modeltime)
library(slider)
library(lubridate)

m4_monthly_extended <- m4_monthly %>%
    group_by(id) %>%
    future_frame(.length_out = 24, .bind_data = TRUE) %>%
    mutate(lag_24 = lag(value, 24)) %>%
    ungroup() %>%
    mutate(date_num = as.numeric(date)) %>%
    mutate(date_month = month(date))


m4_monthly_train  <- m4_monthly_extended %>% drop_na

m4_monthly_future <- m4_monthly_extended %>% filter(is.na(value))

splits <- time_series_split(m4_monthly_train, assess = 24, cumulative = TRUE)

model_fit_gam <- gen_additive_mod(mode = "regression", smoother_dim_term1 = 3, 
                                  smoother_dim_term2 = 3, smoother_dim_term3 = 6) %>%
    set_engine("gam", method = "REML") %>%
    fit(value ~ s(date_month, by = id) 
        + s(date_num, by = id) 
        + s(date_num, date_month, by = id)
        + id,
        data = training(splits))

model_fit_gam

calibration_tbl <- modeltime_table(
    model_fit_gam
) %>%
    modeltime_calibrate(testing(splits), quiet = FALSE) 

calibration_tbl %>% modeltime_accuracy()

calibration_tbl %>%
    modeltime_forecast(
        new_data    = testing(splits),
        actual_data = m4_monthly_train,
        keep_data   = TRUE
    ) %>%
    group_by(id) %>%
    plot_modeltime_forecast(.facet_ncol = 2)

calibration_tbl %>%
    modeltime_refit(m4_monthly_train) %>%
    modeltime_forecast(
        new_data    = m4_monthly_future,
        actual_data = m4_monthly_train,
        keep_data   = TRUE 
    ) %>%
    group_by(id) %>%
    plot_modeltime_forecast()