library(tidyverse)
## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr 1.1.4 ✔ readr 2.1.5
## ✔ forcats 1.0.0 ✔ stringr 1.5.1
## ✔ ggplot2 3.5.1 ✔ tibble 3.2.1
## ✔ lubridate 1.9.3 ✔ tidyr 1.3.1
## ✔ purrr 1.0.2
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag() masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
library(correlationfunnel)
## Warning: package 'correlationfunnel' was built under R version 4.4.2
## ══ Using correlationfunnel? ════════════════════════════════════════════════════
## You might also be interested in applied data science training for business.
## </> Learn more at - www.business-science.io </>
library(tidymodels)
## Warning: package 'tidymodels' was built under R version 4.4.2
## ── Attaching packages ────────────────────────────────────── tidymodels 1.2.0 ──
## ✔ broom 1.0.6 ✔ rsample 1.2.1
## ✔ dials 1.3.0 ✔ tune 1.2.1
## ✔ infer 1.0.7 ✔ workflows 1.1.4
## ✔ modeldata 1.4.0 ✔ workflowsets 1.1.0
## ✔ parsnip 1.2.1 ✔ yardstick 1.3.2
## ✔ recipes 1.1.0
## Warning: package 'dials' was built under R version 4.4.2
## Warning: package 'infer' was built under R version 4.4.2
## Warning: package 'modeldata' was built under R version 4.4.2
## Warning: package 'parsnip' was built under R version 4.4.2
## Warning: package 'tune' was built under R version 4.4.2
## Warning: package 'workflows' was built under R version 4.4.2
## Warning: package 'workflowsets' was built under R version 4.4.2
## Warning: package 'yardstick' was built under R version 4.4.2
## ── Conflicts ───────────────────────────────────────── tidymodels_conflicts() ──
## ✖ scales::discard() masks purrr::discard()
## ✖ dplyr::filter() masks stats::filter()
## ✖ recipes::fixed() masks stringr::fixed()
## ✖ dplyr::lag() masks stats::lag()
## ✖ yardstick::spec() masks readr::spec()
## ✖ recipes::step() masks stats::step()
## • Use tidymodels_prefer() to resolve common conflicts.
library(themis)
## Warning: package 'themis' was built under R version 4.4.3
library(vip)
## Warning: package 'vip' was built under R version 4.4.3
##
## Attaching package: 'vip'
##
## The following object is masked from 'package:utils':
##
## vi
spam <- readr::read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/main/data/2023/2023-08-15/spam.csv')
## Rows: 4601 Columns: 7
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (1): yesno
## dbl (6): crl.tot, dollar, bang, money, n000, make
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.
spam_clean <- spam %>%
mutate(yesno = factor(yesno, levels = c("y", "n")))
spam_clean %>% count(yesno)
## # A tibble: 2 × 2
## yesno n
## <fct> <int>
## 1 y 1813
## 2 n 2788
ggplot(spam_clean, aes(yesno)) + geom_bar()
ggplot(spam_clean, aes(yesno, crl.tot)) + geom_boxplot()
spam_binarized <- spam_clean %>% binarize()
spam_correlation <- spam_binarized %>% correlate(yesno__y)
spam_correlation_sorted <- spam_correlation %>% arrange(desc(correlation))
correlationfunnel::plot_correlation_funnel(spam_correlation_sorted)
set.seed(1234)
spam_split <- initial_split(spam_clean, strata = yesno)
spam_train <- training(spam_split)
spam_test <- testing(spam_split)
spam_cv <- vfold_cv(spam_train, strata = yesno)
spam_recipe <- recipe(yesno ~ ., data = spam_train) %>%
step_YeoJohnson(all_numeric_predictors()) %>%
step_normalize(all_numeric_predictors()) %>%
step_dummy(all_nominal_predictors()) %>%
step_smote(yesno)
spam_xgboost_spec <- boost_tree(trees = tune(), tree_depth = tune()) %>%
set_mode("classification") %>%
set_engine("xgboost")
spam_rf_spec <- rand_forest(trees = tune()) %>%
set_mode("classification") %>%
set_engine("ranger", importance = "impurity")
spam_xgboost_workflow <- workflow() %>% add_recipe(spam_recipe) %>% add_model(spam_xgboost_spec)
spam_rf_workflow <- workflow() %>% add_recipe(spam_recipe) %>% add_model(spam_rf_spec)
doParallel::registerDoParallel()
xgboost_grid <- grid_regular(trees(), tree_depth(), levels = 5)
rf_grid <- grid_regular(trees(), levels = 5)
set.seed(43931)
spam_xgboost_tune <- tune_grid(spam_xgboost_workflow, resamples = spam_cv, grid = xgboost_grid, control = control_grid(save_pred = TRUE))
spam_rf_tune <- tune_grid(spam_rf_workflow, resamples = spam_cv, grid = rf_grid, control = control_grid(save_pred = TRUE))
xgboost_metrics <- collect_metrics(spam_xgboost_tune)
rf_metrics <- collect_metrics(spam_rf_tune)
xgboost_roc <- collect_predictions(spam_xgboost_tune) %>% group_by(id) %>% roc_curve(yesno, .pred_y) %>% autoplot()
rf_roc <- collect_predictions(spam_rf_tune) %>% group_by(id) %>% roc_curve(yesno, .pred_y) %>% autoplot()
list(xgboost_roc, rf_roc)
## [[1]]
##
## [[2]]
spam_xgboost_final <- spam_xgboost_workflow %>% finalize_workflow(select_best(spam_xgboost_tune, metric = "accuracy")) %>% last_fit(spam_split)
## Warning: package 'xgboost' was built under R version 4.4.2
spam_rf_final <- spam_rf_workflow %>% finalize_workflow(select_best(spam_rf_tune, metric = "accuracy")) %>% last_fit(spam_split)
## Warning: package 'ranger' was built under R version 4.4.2
final_metrics <- bind_rows(collect_metrics(spam_xgboost_final), collect_metrics(spam_rf_final))
spam_xgboost_final %>% workflows::extract_fit_engine() %>% vip()
spam_rf_final %>% workflows::extract_fit_engine() %>% vip()
The initial model used YeoJohnson transformation. Enhancements included: * Normalization of numeric features. * SMOTE to balance the target variable. * Hyperparameter tuning for tree depth and number of trees. * Random Forest model comparison.
The final results showed: * XGBoost achieved an accuracy of 0.858 and ROC AUC of 0.914. * Random Forest achieved an accuracy of 0.880 and ROC AUC of 0.921.
This comparison provides insight into which model performs better for spam detection.