好みの日本酒を呑みたい！ ?さけのわデータで探す自分好みの酒??
2 likes6,567 views
This document analyzes Japanese sake brand data to cluster brands and identify similarities. It first calculates cosine distances between brands based on flavor profiles. It then performs hierarchical clustering on the distances to group brands into 6 clusters. Finally, it creates radar charts of average flavor profiles and word clouds of common tags for each cluster to visualize differences between the groups.
![> me
$name
[1] "Takashi Kitano"
$twitter
[1] "@kashitan"
$work_in
[1] " "](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-2-320.jpg)
![brands.dist.mat %>%
as.matrix() %>%
.[1:8, 1:8]](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-9-320.jpg)
![radars <-
radars %>%
#
dplyr::group_by(cluster) %>%
tidyr::nest() %>%
dplyr::mutate(fig = purrr::map2(data, cluster, function(x, y) {
plotly::plot_ly(
type = "scatterpolar", mode = "markers",
r = c(x$f1, x$f2, x$f3, x$f4, x$f5, x$f6, x$f1),
theta = c(" ", " ", " ", " ", " ", " ", " "),
fill = 'toself',
fillcolor = RColorBrewer::brewer.pal(n = n, name = "Accent")[y],
opacity = 0.5
) %>%
plotly::layout(polar = list(angularaxis = list(
rotation = 90,
direction = 'counterclockwise')))
}))](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-18-320.jpg)
![radars$fig[[1]]](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-19-320.jpg)
![radars$fig[[4]]](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-20-320.jpg)
![radars$fig[[2]]](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-21-320.jpg)
![radars$fig[[6]]](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-22-320.jpg)
![radars$fig[[3]]](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-23-320.jpg)
![radars$fig[[5]]](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-24-320.jpg)
![brands.mds <-
brands.dist.mat %>%
# (MDS)
cmdscale() %>%
`colnames<-`(c("x", "y")) %>%
tibble::as_tibble(rownames = NA) %>%
tibble::rownames_to_column(var = "brandId") %>%
dplyr::mutate(brandId = as.integer(brandId)) %>%
#
dplyr::inner_join(
brands.cluster,
by = c("brandId" = "brandId")
) %>%
#
dplyr::inner_join(
brands[, -3],
by = c("brandId" = "id")
) %>%
dplyr::mutate(cluster = forcats::as_factor(cluster))
1](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-25-320.jpg)
![#
res.ca <-
FactoMineR::CA(contingency.table, graph = FALSE)
# tibble
tags.biplot <-
tibble::tibble(
type = "tag",
x = res.ca$row$coord[, 1],
y = res.ca$row$coord[, 2],
label = rownames(contingency.table)) %>%
dplyr::bind_rows(
tibble::tibble(
type = "cluster",
x = res.ca$col$coord[, 1],
y = res.ca$col$coord[, 2],
label = colnames(contingency.table))
)](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-38-320.jpg)
![tags.biplot %>%
plotly::plot_ly(x =~x, y =~y) %>%
plotly::add_markers(color = ~type,
colors = RColorBrewer::brewer.pal(3, "Set1")[1:2]) %>%
plotly::add_text(text = ~label, textposition = "top center") %>%
plotly::layout(showlegend = FALSE)
factoextra::fviz_ca_biplot(
res.ca,
font.family = "HiraKakuProN-W3"
)](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-39-320.jpg)
![radars$fig[[1]] wordclouds$fig[[1]]](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-43-320.jpg)
![radars$fig[[4]] wordclouds$fig[[4]]](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-44-320.jpg)
![radars$fig[[2]] wordclouds$fig[[2]]](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-45-320.jpg)
![radars$fig[[6]] wordclouds$fig[[6]]](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-46-320.jpg)
![radars$fig[[3]] wordclouds$fig[[3]]](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-47-320.jpg)
![radars$fig[[5]] wordclouds$fig[[5]]](https://image.slidesharecdn.com/20200919tokyor88-200919060725/85/-48-320.jpg)
More Related Content
What's hot (20)
Similar to 好みの日本酒を呑みたい！ ?さけのわデータで探す自分好みの酒?? (20)
More from Takashi Kitano (11)
Recently uploaded (20)
好みの日本酒を呑みたい！ ?さけのわデータで探す自分好みの酒??
- 1. ?
- 2. > me $name [1] "Takashi Kitano" $twitter [1] "@kashitan" $work_in [1] " "
- 3. ? ?
- 6. ## charts <- jsonlite::read_json("https://muro.sakenowa.com/sakenowa-data/api/flavor-charts", simplifyVector = TRUE) %>% purrr::pluck("flavorCharts") ↑ ↑ ↑ ↑ ↑ ↑
- 7. ## brands <- jsonlite::read_json("https://muro.sakenowa.com/sakenowa-data/api/brands", simplifyVector = TRUE) %>% purrr::pluck("brands")
- 8. ## brands.dist.mat <- charts %>% # tibble::remove_rownames() %>% # ID tibble::column_to_rownames(var = "brandId") %>% # proxy::dist(method = "cosine")
- 10. ## data.frame brands.dist.df <- brands.dist.mat %>% as.matrix() %>% # tibble tibble::as_tibble(rownames = NA) %>% # tibble::rownames_to_column(var = "brandId1") %>% # tidyr::pivot_longer( cols = -brandId1, names_to = "brandId2", values_to = "dist") %>% # dplyr::filter(brandId1 != brandId2) %>% # ID dplyr::mutate_if(is.character, as.integer)
- 11. brands.dist.df %>% # dplyr::inner_join(brands, by = c("brandId1" = "id")) %>% dplyr::inner_join(brands, by = c("brandId2" = "id")) %>% # dplyr::filter(name.x == " ") %>% # dplyr::arrange(dist) %>% dplyr::select(name.y, dist)
- 13. brands.dist.df %>% # dplyr::inner_join(brands, by = c("brandId1" = "id")) %>% dplyr::inner_join(brands, by = c("brandId2" = "id")) %>% # dplyr::filter(name.x == " ") %>% # dplyr::arrange(desc(dist)) %>% dplyr::select(name.y, dist)
- 16. n <- 6 # ( 6 ) brands.cluster <- brands.dist.mat %>% # hclust(method="ward.D2") %>% # cutree(n) %>% # tibble tibble::tibble( brandId = as.integer(names(.)), cluster = . )
- 17. radars <- brands.cluster %>% # dplyr::inner_join(charts, by = c("brandId" = "brandId")) %>% # dplyr::group_by(cluster) %>% dplyr::summarise_at(dplyr::vars(dplyr::starts_with("f")), mean) 1
- 18. radars <- radars %>% # dplyr::group_by(cluster) %>% tidyr::nest() %>% dplyr::mutate(fig = purrr::map2(data, cluster, function(x, y) { plotly::plot_ly( type = "scatterpolar", mode = "markers", r = c(x$f1, x$f2, x$f3, x$f4, x$f5, x$f6, x$f1), theta = c(" ", " ", " ", " ", " ", " ", " "), fill = 'toself', fillcolor = RColorBrewer::brewer.pal(n = n, name = "Accent")[y], opacity = 0.5 ) %>% plotly::layout(polar = list(angularaxis = list( rotation = 90, direction = 'counterclockwise'))) }))
- 19. radars$fig[[1]]
- 20. radars$fig[[4]]
- 21. radars$fig[[2]]
- 22. radars$fig[[6]]
- 23. radars$fig[[3]]
- 24. radars$fig[[5]]
- 25. brands.mds <- brands.dist.mat %>% # (MDS) cmdscale() %>% `colnames<-`(c("x", "y")) %>% tibble::as_tibble(rownames = NA) %>% tibble::rownames_to_column(var = "brandId") %>% dplyr::mutate(brandId = as.integer(brandId)) %>% # dplyr::inner_join( brands.cluster, by = c("brandId" = "brandId") ) %>% # dplyr::inner_join( brands[, -3], by = c("brandId" = "id") ) %>% dplyr::mutate(cluster = forcats::as_factor(cluster)) 1
- 27. brands.mds %>% # dplyr::filter(brandId %in% rankings$overall$brandId) %>% # plotly::plot_ly(x = ~x, y = ~y) %>% plotly::add_markers( color = ~cluster, text = ~name, colors = RColorBrewer::brewer.pal(n = n, name = "Accent")) %>% plotly::add_text( text = ~name, textposition = "top center" ) %>% plotly::layout(showlegend = FALSE)
- 34. ## tags <- jsonlite::read_json("https://muro.sakenowa.com/sakenowa-data/api/flavor-tags", simplifyVector = TRUE) %>% purrr::pluck("tags")
- 35. ## brand_tags <- jsonlite::read_json( "https://muro.sakenowa.com/sakenowa-data/api/brand-flavor-tags", simplifyVector = TRUE ) %>% purrr::pluck("flavorTags")
- 36. # contingency.table <- brand_tags %>% tidyr::unnest(cols = tagIds) %>% # dplyr::inner_join( brands.cluster, by = c("brandId" = "brandId") ) %>% # dplyr::inner_join( tags, by = c("tagIds" = "id") ) %>% dplyr::group_by(tagIds, tag, cluster) %>% dplyr::count() 1
- 37. contingency.table <- contingency.table %>% # tidyr::pivot_wider( id_cols = tag, names_from = cluster, values_from = n, values_fill = 0) %>% # tibble::column_to_rownames(var = "tag")
- 38. # res.ca <- FactoMineR::CA(contingency.table, graph = FALSE) # tibble tags.biplot <- tibble::tibble( type = "tag", x = res.ca$row$coord[, 1], y = res.ca$row$coord[, 2], label = rownames(contingency.table)) %>% dplyr::bind_rows( tibble::tibble( type = "cluster", x = res.ca$col$coord[, 1], y = res.ca$col$coord[, 2], label = colnames(contingency.table)) )
- 39. tags.biplot %>% plotly::plot_ly(x =~x, y =~y) %>% plotly::add_markers(color = ~type, colors = RColorBrewer::brewer.pal(3, "Set1")[1:2]) %>% plotly::add_text(text = ~label, textposition = "top center") %>% plotly::layout(showlegend = FALSE) factoextra::fviz_ca_biplot( res.ca, font.family = "HiraKakuProN-W3" )
- 42. wordclouds <- brand_tags %>% tidyr::unnest(cols = tagIds) %>% dplyr::inner_join( brands.cluster, by = c("brandId" = "brandId") ) %>% dplyr::inner_join(tags, by = c("tagIds" = "id")) %>% # dplyr::group_by(cluster, tag) %>% dplyr::count() %>% # dplyr::group_by(cluster) %>% dplyr::arrange(cluster, desc(n)) %>% tidyr::nest() %>% dplyr::mutate(fig = purrr::map( data, wordcloud2::wordcloud2, size = 0.8, minSize = 10 ))