Prognoza MSCI Europe Index Anunț tarifar post-Trump

Stocurile europene cresc brusc după ce Trump amânează tariful de 50% pentru UE. Potrivit modelului ML, acest lucru poate crește prețurile la trupa de 18.000 de dolari pentru indicele MSCI Europe.

library(tidymodels)
library(tidyverse)
library(tidyquant)
library(modeltime)
library(timetk)

#MSCI Europe Index
#download:https://www.msci.com/indexes/index/990500
df_msci_eur <- 
  read_csv("data/msci_eur.csv") %>% 
  janitor::clean_names() %>% 
  rename(close = value)

# Split Data 
splits <- time_series_split(df_msci_eur,
                            assess = "3 months",
                            cumulative = TRUE)

df_train <- training(splits)
df_test <- testing(splits)

#Model 1: Auto ARIMA
model_fit_arima_no_boost <- 
  arima_reg() %>%
  set_engine(engine = "auto_arima") %>%
  fit(close ~ date, data = df_train)


#Model 2: Boosted Auto ARIMA
model_fit_arima_boosted <- 
  arima_boost(
    min_n = 2,
    learn_rate = 0.015
  ) %>%
  set_engine(engine = "auto_arima_xgboost") %>%
  fit(close ~ date + as.numeric(date) + factor(lubridate::month(date, label = TRUE), 
                                               ordered = F),
      data = df_train)


#Model 3: Exponential Smoothing
model_fit_ets <- 
  exp_smoothing() %>%
  set_engine(engine = "ets") %>%
  fit(close ~ date, data = df_train)  


#Model 4: Prophet
model_fit_prophet <- 
  prophet_reg() %>%
  set_engine(engine = "prophet") %>%
  fit(close ~ date, data = df_train)  


#Model 5: Linear Regression
model_fit_lm <- 
  linear_reg() %>%
  set_engine("lm") %>%
  fit(close ~ as.numeric(date) + factor(lubridate::month(date, label = TRUE), 
                                        ordered = FALSE),
      data = df_train)


#Model 6: MARS
library(earth)

model_spec_mars <- 
  mars(mode = "regression") %>%
  set_engine("earth") 

recipe_spec <- 
  recipe(close ~ date, data = df_train) %>%
  step_date(date, features = "month", ordinal = FALSE) %>%
  step_mutate(date_num = as.numeric(date)) %>%
  step_normalize(date_num) %>%
  step_rm(date)

wflw_fit_mars <- 
  workflow() %>%
  add_recipe(recipe_spec) %>%
  add_model(model_spec_mars) %>%
  fit(df_train)

#Radial basis function support vector machines
model_spec_svm <- 
  svm_rbf(cost = 1, margin = 0.1) %>% 
  set_mode("regression") %>% 
  set_engine("kernlab")

wflw_fit_svm <- 
  workflow() %>%
  add_recipe(recipe_spec) %>%
  add_model(model_spec_svm) %>%
  fit(df_train)


#Model Tables
models_tbl <- modeltime_table(
  model_fit_arima_no_boost,
  model_fit_arima_boosted,
  model_fit_ets,
  model_fit_prophet,
  model_fit_lm,
  wflw_fit_mars,
  wflw_fit_svm
)


#Calibration
calibration_tbl <-
  models_tbl %>%
  modeltime_calibrate(new_data = df_test)

#Accuracy
calibration_tbl %>%
  modeltime_accuracy() %>%
  table_modeltime_accuracy(
    .interactive = TRUE
  )


#Calibrate the Linear Regression model to a testing set
calibration_lm <- 
  model_fit_lm %>%
  modeltime_calibrate(new_data = df_test)


#Predictive intervals (95% Confidence Interval)
calibration_lm %>%
  modeltime_forecast(actual_data = df_msci_eur %>% 
                                   filter(date >= last(date) - months(3)),
                     new_data = testing(splits)) %>% 
  plot_modeltime_forecast(.interactive = FALSE,
                          .legend_show = FALSE,
                          .line_size = 1.5,
                          .color_lab = "",
                          .title = "MSCI Europe Index") +
  ggrepel::geom_text_repel(
    data = df_msci_eur %>%  slice_tail(n = 1),
    aes(label = paste0("$",format(round(close , 2), big.mark = ",")), 
        x= date,
        y = close),
    hjust = 1, 
    vjust = .5,
    nudge_x = 0.5,
    fontface = "bold", 
    family = "Roboto Slab",
    size = 6,
    color = "black",
    segment.color = NA
  ) +
  ggrepel::geom_text_repel(
    data = . %>%  slice_tail(n = 1),
    aes(label = paste0("$",format(round(.conf_hi,2), big.mark = ",")), 
        x= .index,
        y = .conf_hi),
    hjust = 1, 
    vjust = 1,
    nudge_x = .5,
    fontface = "bold", 
    family = "Roboto Slab",
    size = 6,
    color = "dimgray",
    segment.color = NA
  ) +
  labs(subtitle = "Predictive Intervals
ML Model") + 
  scale_x_date(expand = expansion(mult = c(.1, .25)),
               labels = scales::label_date(format = "%b'%y")) +
  scale_y_continuous(labels = scales::label_currency()) +
  theme_minimal(base_family = "Roboto Slab", base_size = 20) +
   theme(legend.position = "none",
        plot.background = element_rect(fill = "azure", 
                                       color = "azure"),
        plot.title = element_text(face = "bold"),
        axis.text = element_text(face = "bold"),
        plot.subtitle = ggtext::element_markdown(face = "bold"))
################################################################################