OLIST02：多重軸度比較圖

---

pacman::p_load(dplyr, ggplot2, readr, plotly, googleVis)

load("data/olist.rdata")
load("data/Z.rdata")

---

A. 基本繪圖套件 ggplot

ggplot(segment, aes(x=log(avgItemsSold), y=avgPrice, col=avgScore)) +
  geom_point(aes(size=sqrt(noSellers))) +
  geom_text(aes(label=business_segment), size=3)

---

B. 互動式繪圖套件 ggplotly

g = ggplot(segment, aes(
  x=log(avgItemsSold), y=avgPrice, col=avgScore, label=business_segment)) +
  geom_point(aes(size=sqrt(noSellers)))
ggplotly(g)

---

C. 互動式多軸度泡泡圖 googleVis

最簡單的做法

# op = options(gvis.plot.tag='chart')
segment$year = 2018
gvisMotionChart(segment, "business_segment", "year") %>% plot

稍微複雜的案例

Merge TPC$product_category_name_english into P as P$category

P = left_join(P, TPC) %>% rename(category = product_category_name_english)

Joining, by = "product_category_name"

Merge P$category intoI`

I = left_join(I, P[,c(1,10)])

Joining, by = "product_id"

Summerise by category

category = filter(I, !is.na(category)) %>% 
  group_by(category) %>% summarise(
    itemsSold = n(),
    totalRev = sum(price),
    avgPrice = mean(price),
    noProduct = n_distinct(product_id),
    noSeller = n_distinct(seller_id),
    dummy = 2018
  ) %>% arrange(desc(totalRev))

`summarise()` ungrouping output (override with `.groups` argument)

Top-20 categories

top20 = category %>% top_n(20, totalRev)
top20

# A tibble: 20 x 7
   category              itemsSold totalRev avgPrice noProduct noSeller dummy
   <chr>                     <int>    <dbl>    <dbl>     <int>    <int> <dbl>
 1 health_beauty              9670 1258681.    130.       2444      492  2018
 2 watches_gifts              5991 1205006.    201.       1329      101  2018
 3 bed_bath_table            11115 1036989.     93.3      3029      196  2018
 4 sports_leisure             8641  988049.    114.       2867      481  2018
 5 computers_accessories      7827  911954.    117.       1639      287  2018
 6 furniture_decor            8334  729762.     87.6      2657      370  2018
 7 cool_stuff                 3796  635291.    167.        789      267  2018
 8 housewares                 6964  632249.     90.8      2335      468  2018
 9 auto                       4235  592720.    140.       1900      383  2018
10 garden_tools               4347  485256.    112.        753      237  2018
11 toys                       4117  483947.    118.       1411      252  2018
12 baby                       3065  411765.    134.        919      244  2018
13 perfumery                  3419  399125.    117.        868      175  2018
14 telephony                  4545  323668.     71.2      1134      149  2018
15 office_furniture           1691  273961.    162.        309       34  2018
16 stationery                 2517  230943.     91.8       849      173  2018
17 computers                   203  222963.   1098.         30        9  2018
18 pet_shop                   1947  214315.    110.        719      137  2018
19 musical_instruments         680  191499.    282.        289       70  2018
20 small_appliances            679  190649.    281.        231      105  2018

靜態多軸互動

gvisMotionChart(category, "category", "dummy") %>% plot

時間面板資料 Panel Data

併入時間資料

X = left_join(O[, c(1,4)], R[,2:3]) %>%     # pull score & timestamp into 'O'
  rename(
    time = order_purchase_timestamp, 
    score = review_score) %>% 
  mutate(                                   # cut timestamp into quarter    
    quarter = as.Date(cut(time, "quarter"))
    ) %>%  
  right_join(I) %>%                         # merge score & quarter into 'I'
  filter(category %in% top20$category) %>%  # pick out the top20 categories
  group_by(category, quarter) %>% 
  summarise(                            # summarise by category & quarter
    itemsSold = n(),                     
    totalRev = sum(price),
    avgPrice = mean(price),
    avgScore = mean(score),
    noProduct = n_distinct(product_id),
    noSeller = n_distinct(seller_id)
  ) %>% 
  arrange(category, quarter)            # order by category & quarter

Joining, by = "order_id"
Joining, by = "order_id"

`summarise()` regrouping output by 'category' (override with `.groups` argument)

調整資料範圍、去除離群值

X2 = X %>%  # adjustment before ploting
  filter(quarter >= as.Date("2017-04-01")) %>% 
  filter(!(category %in% c("computers", "office_furniture"))) %>% 
  mutate(avgPrice = pmax(avgPrice, 3)) %>% as.data.frame

動態多軸互動

gvisMotionChart(X2, "category", "quarter") %>% plot

---