Apply 2
Ronja Dahlin
2025-02-27
Import Data
ikea <- read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2020/2020-11-03/ikea.csv')## New names:
## Rows: 3694 Columns: 14
## ── Column specification
## ──────────────────────────────────────────────────────── Delimiter: "," chr
## (7): name, category, old_price, link, other_colors, short_description, d... dbl
## (6): ...1, item_id, price, depth, height, width lgl (1): sellable_online
## 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.
## • `` -> `...1` skimr::skim(ikea)| Name | ikea |
| Number of rows | 3694 |
| Number of columns | 14 |
| _______________________ | |
| Column type frequency: | |
| character | 7 |
| logical | 1 |
| numeric | 6 |
| ________________________ | |
| Group variables | None |
Variable type: character
| skim_variable | n_missing | complete_rate | min | max | empty | n_unique | whitespace |
|---|---|---|---|---|---|---|---|
| name | 0 | 1 | 3 | 27 | 0 | 607 | 0 |
| category | 0 | 1 | 4 | 36 | 0 | 17 | 0 |
| old_price | 0 | 1 | 4 | 13 | 0 | 365 | 0 |
| link | 0 | 1 | 52 | 163 | 0 | 2962 | 0 |
| other_colors | 0 | 1 | 2 | 3 | 0 | 2 | 0 |
| short_description | 0 | 1 | 3 | 63 | 0 | 1706 | 0 |
| designer | 0 | 1 | 3 | 1261 | 0 | 381 | 0 |
Variable type: logical
| skim_variable | n_missing | complete_rate | mean | count |
|---|---|---|---|---|
| sellable_online | 0 | 1 | 0.99 | TRU: 3666, FAL: 28 |
Variable type: numeric
| skim_variable | n_missing | complete_rate | mean | sd | p0 | p25 | p50 | p75 | p100 | hist |
|---|---|---|---|---|---|---|---|---|---|---|
| …1 | 0 | 1.00 | 1846.50 | 1066.51 | 0 | 923.25 | 1846.5 | 2769.75 | 3693 | ▇▇▇▇▇ |
| item_id | 0 | 1.00 | 48632396.79 | 28887094.10 | 58487 | 20390574.00 | 49288078.0 | 70403572.75 | 99932615 | ▇▇▇▇▇ |
| price | 0 | 1.00 | 1078.21 | 1374.65 | 3 | 180.90 | 544.7 | 1429.50 | 9585 | ▇▁▁▁▁ |
| depth | 1463 | 0.60 | 54.38 | 29.96 | 1 | 38.00 | 47.0 | 60.00 | 257 | ▇▃▁▁▁ |
| height | 988 | 0.73 | 101.68 | 61.10 | 1 | 67.00 | 83.0 | 124.00 | 700 | ▇▂▁▁▁ |
| width | 589 | 0.84 | 104.47 | 71.13 | 1 | 60.00 | 80.0 | 140.00 | 420 | ▇▅▂▁▁ |
data <- ikea %>%
na.omit() %>%
# Remove unnecessary columns and treat missing values
select(-old_price, -link, -...1) %>%
# Handle multiple designers
separate_rows(designer, sep = "/") %>%
# Convert character variables to factors
mutate(across(where(is.character), as.factor)) %>%
mutate(short_description = as.character(short_description)) %>%
# Convert logical variables to factors
mutate(across(where(is.logical), as.factor)) %>%
# Log transform price (ensure no zero or negative values)
mutate(price = log(price))Explore Data
Identify good predictors
# Category
data %>%
ggplot(aes(price, category)) +
geom_boxplot()
# Designer
data %>%
# Tokenize description text
unnest_tokens(output = word, input = short_description) %>%
# calculate avg rent per word
group_by(word) %>%
summarise(price = mean(price),
n = n()) %>%
ungroup() %>%
filter(n > 10, !str_detect(word, "\\d")) %>%
slice_max(order_by = price, n = 10) %>%
# Plot
ggplot(aes(price, fct_reorder(word, price))) +
geom_point() +
labs(y = "Products")
EDA shortcut
# Step 1:Prepare data
#Select
#take out short description, and keep designer
data_binarize_tbl <- data %>%
select(price, height, width, depth, category, other_colors, sellable_online) %>%
binarize()
data_binarize_tbl %>% glimpse()## Rows: 2,714
## Columns: 35
## $ `price__-Inf_5.84354441703136` <dbl> 1, 1, 1, 1, 1, 1, 1, 1, 1…
## $ price__5.84354441703136_6.75693238924755 <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ price__6.75693238924755_7.52023455647463 <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ price__7.52023455647463_Inf <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ `height__-Inf_74` <dbl> 0, 1, 0, 0, 0, 0, 0, 0, 0…
## $ height__74_95 <dbl> 0, 0, 1, 1, 1, 1, 1, 0, 0…
## $ height__95_180.75 <dbl> 1, 0, 0, 0, 0, 0, 0, 1, 1…
## $ height__180.75_Inf <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ `width__-Inf_60` <dbl> 1, 0, 1, 1, 1, 1, 1, 1, 1…
## $ width__60_100 <dbl> 0, 1, 0, 0, 0, 0, 0, 0, 0…
## $ width__100_180 <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ width__180_Inf <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ `depth__-Inf_40` <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ depth__40_48 <dbl> 0, 0, 1, 1, 1, 1, 1, 1, 1…
## $ depth__48_66 <dbl> 1, 1, 0, 0, 0, 0, 0, 0, 0…
## $ depth__66_Inf <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ category__Bar_furniture <dbl> 1, 1, 1, 1, 1, 1, 1, 1, 1…
## $ category__Beds <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ `category__Bookcases_&_shelving_units` <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ `category__Cabinets_&_cupboards` <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ category__Café_furniture <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ category__Chairs <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ `category__Chests_of_drawers_&_drawer_units` <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ `category__Children's_furniture` <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ category__Nursery_furniture <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ category__Outdoor_furniture <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ `category__Sofas_&_armchairs` <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ `category__Tables_&_desks` <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ `category__TV_&_media_furniture` <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ category__Wardrobes <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ `category__-OTHER` <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ other_colors__No <dbl> 0, 1, 1, 1, 1, 1, 1, 1, 1…
## $ other_colors__Yes <dbl> 1, 0, 0, 0, 0, 0, 0, 0, 0…
## $ sellable_online__TRUE <dbl> 1, 1, 1, 1, 1, 1, 1, 1, 1…
## $ `sellable_online__-OTHER` <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0…# Step 2: Correlate
data_corr_tbl <- data_binarize_tbl %>%
correlate(price__7.52023455647463_Inf)
data_corr_tbl## # A tibble: 35 × 3
## feature bin correlation
## <fct> <chr> <dbl>
## 1 price 7.52023455647463_Inf 1
## 2 width 180_Inf 0.610
## 3 depth 66_Inf 0.418
## 4 width -Inf_60 -0.356
## 5 category Sofas_&_armchairs 0.343
## 6 price -Inf_5.84354441703136 -0.335
## 7 price 6.75693238924755_7.52023455647463 -0.333
## 8 price 5.84354441703136_6.75693238924755 -0.332
## 9 category Wardrobes 0.302
## 10 height -Inf_74 -0.283
## # ℹ 25 more rows# Step 3: Plot
data_corr_tbl %>%
plot_correlation_funnel() ## Warning: ggrepel: 12 unlabeled data points (too many overlaps). Consider
## increasing max.overlaps
Build Models
Split data
data2 <- data %>%
select(-other_colors)
# Split into train and test data set
set.seed(1234)
ikea_split <- rsample::initial_split(data2, prop = 0.75)
ikea_train <- training(ikea_split)
ikea_test <- testing(ikea_split)# Specify recipe
ikea_recipe <- recipe(formula = price ~ ., data = ikea_train) %>%
update_role(item_id, new_role = "id variable") %>%
step_tokenize(short_description) %>%
step_tokenfilter(short_description, max_tokens = 10) %>% # Missing argument
step_tfidf(short_description) %>%
step_other(category, name, sellable_online, designer, threshold = 0.05) %>%
step_dummy(category, name, sellable_online, designer, one_hot = TRUE) %>%
step_YeoJohnson(height, width, depth) %>%
step_impute_knn(all_numeric_predictors()) # Fixes the broken pipeline
# Now, prep and juice
ikea_recipe %>% prep() %>% juice() %>% glimpse() ## Rows: 2,035
## Columns: 32
## $ item_id <dbl> 9219527, 99248380, 59157555,…
## $ depth <dbl> 4.816575, 4.055862, 4.896875…
## $ height <dbl> 13.192531, 15.758387, 20.498…
## $ width <dbl> 5.949197, 10.244117, 8.93084…
## $ price <dbl> 4.595120, 6.897705, 7.338238…
## $ tfidf_short_description_3 <dbl> 0.000000, 1.278636, 1.278636…
## $ tfidf_short_description_bed <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ tfidf_short_description_cm <dbl> 0.0000000, 0.4441684, 0.4441…
## $ tfidf_short_description_combination <dbl> 0.0000000, 0.0000000, 0.0000…
## $ tfidf_short_description_doors <dbl> 0.0000000, 0.0000000, 0.0000…
## $ tfidf_short_description_seat <dbl> 0.0000000, 0.0000000, 0.0000…
## $ tfidf_short_description_sofa <dbl> 0.000000, 0.000000, 0.000000…
## $ tfidf_short_description_storage <dbl> 0.0000000, 0.0000000, 0.0000…
## $ tfidf_short_description_wardrobe <dbl> 0.000000, 0.000000, 0.000000…
## $ tfidf_short_description_with <dbl> 0.0000000, 0.0000000, 0.0000…
## $ category_Beds <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ category_Bookcases...shelving.units <dbl> 0, 1, 0, 0, 1, 0, 0, 0, 0, 1…
## $ category_Cabinets...cupboards <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ category_Chairs <dbl> 1, 0, 0, 0, 0, 0, 0, 0, 1, 0…
## $ category_Chests.of.drawers...drawer.units <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ category_Sofas...armchairs <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ category_Wardrobes <dbl> 0, 0, 1, 0, 0, 0, 0, 1, 0, 0…
## $ category_other <dbl> 0, 0, 0, 1, 0, 1, 1, 0, 0, 0…
## $ name_BESTÅ <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ name_PAX <dbl> 0, 0, 0, 0, 0, 0, 0, 1, 0, 0…
## $ name_other <dbl> 1, 1, 1, 1, 1, 1, 1, 0, 1, 1…
## $ sellable_online_TRUE. <dbl> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1…
## $ sellable_online_other <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ designer_Ehlén.Johansson <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ designer_IKEA.of.Sweden <dbl> 1, 1, 1, 0, 0, 1, 1, 0, 1, 0…
## $ designer_Ola.Wihlborg <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0…
## $ designer_other <dbl> 0, 0, 0, 1, 1, 0, 0, 1, 0, 1…# Specify model
xgboost_spec <-
boost_tree(trees = tune(), min_n = tune(), mtry = tune(), learn_rate = tune())%>%
set_mode("regression") %>%
set_engine("xgboost")
# Combine recipe and model using workflow
xgboost_workflow <-
workflow() %>%
add_recipe(ikea_recipe) %>%
add_model(xgboost_spec)
# Create cross-validation folds
set.seed(2345)
ikea_cv <- vfold_cv(ikea_train, v = 5)
# Tune hyperperameters
set.seed(3456)
xgboost_tune <-
tune_grid(xgboost_workflow,
resamples = ikea_cv,
grid = 10)## i Creating pre-processing data to finalize unknown parameter: mtry