Uniform Manifold Approximation and Projection (UMAP)

Reduksi Dimensi di R

Matt Pickard

Owner, Pickard Predictives, LLC

PCA, t-SNE, dan UMAP

Perbandingan PCA, t-SNE, dan UMAP

PCA, t-SNE, dan UMAP

Perbandingan PCA, t-SNE, dan UMAP

PCA, t-SNE, dan UMAP

Perbandingan PCA, t-SNE, dan UMAP

PCA, t-SNE, dan UMAP

Perbandingan PCA, t-SNE, dan UMAP

PCA, t-SNE, dan UMAP

Perbandingan PCA, t-SNE, dan UMAP

UMAP memiliki hiperparameter serupa yang dapat dituning.

Plot UMAP

library(embed)


set.seed(1234)
umap_df <- recipe(Attrition ~ ., data = attrition_df) %>% 
  step_normalize(all_predictors()) %>% 
  step_umap(all_predictors(), num_comp = 2) %>% 
  prep() %>% 
  juice()


umap_df %>%  
  ggplot(aes(x = UMAP1, y = UMAP2, color = Attrition)) +
  geom_point(alpha = 0.7)

UMAP: attrition karyawan

Plot UMAP untuk attrition karyawan

UMAP di tidymodels

Buat resep

umap_recipe <-  recipe(Attrition ~ ., data = train) %>% 
  step_normalize(all_predictors()) %>% 
  step_umap(all_predictors(), num_comp = 4)

Buat spesifikasi model

umap_lr_model <- linear_reg()

UMAP di tidymodels

Buat workflow

umap_lr_workflow <-  workflow() %>% 
  add_recipe(umap_recipe) %>% 
  add_model(umap_lr_model)

Fit workflow

umap_lr_fit <- umap_lr_workflow %>% 
  fit(data = train)

UMAP di tidymodels

Evaluasi model

predict_umap_df <- test %>% 
  bind_cols(predict = predict(umap_lr_fit, test))

rmse(predict_umap_df, Attrition, .pred_class)

Ayo berlatih!

Reduksi Dimensi di R