-
Notifications
You must be signed in to change notification settings - Fork 0
/
kmeans 1.R
171 lines (140 loc) · 6.02 KB
/
kmeans 1.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
# https://www.statmethods.net/advstats/cluster.html
# https://uc-r.github.io/kmeans_clustering
# https://www.datanovia.com/en/blog/types-of-clustering-methods-overview-and-quick-start-r-code/
# https://rpkgs.datanovia.com/factoextra/reference/fviz_silhouette.html
## Libraries ----
# Read SPSS
library(foreign)
#General
library(tidyverse)
# Clustering
library(caret)
library(cluster)
library(factoextra)
library(psych)
library(NbClust)
# Visualization
library(skimr)
library(ggplot2)
library(ggpubr)
## Read data ----
data <- read.spss('data/provinces.sav', to.data.frame=TRUE) %>%
mutate(provincia = trimws(provincia)) %>%
drop_na()
skim(data)
## Functions ----
method_number <- function(data, names, centers = 7) {
set.seed(42)
data_limited <- data[, names]
# Elbow method
# https://scikit-learn.org/stable/modules/clustering.html
p_elbow <- fviz_nbclust(data_limited, kmeans, method = "wss") +
geom_vline(xintercept = centers, linetype = 2)+
labs(subtitle = "Elbow method")
# Silhouette method
p_silhouette <- fviz_nbclust(data_limited, kmeans, method = "silhouette", print.summary = TRUE) +
labs(subtitle = "Silhouette method")
# Gap statistic
p_gap_statistic <- fviz_nbclust(data_limited, kmeans, nstart = 25, method = "gap_stat", nboot = 50)+
labs(subtitle = "Gap statistic method")
# NbClust
res.nbclust <- NbClust(data_limited, distance = "euclidean", min.nc = 2, max.nc = 10, method = "complete", index ="all")
# Visualize
p_NbClust <- fviz_nbclust(res.nbclust, ggtheme = theme_minimal())
return(list(
p_elbow = p_elbow,
p_silhouette = p_silhouette,
p_gap_statistic = p_gap_statistic,
p_NbClust = p_NbClust,
p_all = ggarrange(p_elbow, p_silhouette, p_gap_statistic, p_NbClust, nrow = 2, ncol = 2)
))
}
# Visualize silhouhette information
method_calculate <- function(data, names, centers = 2, print.summary = FALSE) {
data_limited <- data[, names]
set.seed(42)
km.res <- kmeans(data_limited, centers, nstart = 25)
sil <- silhouette(km.res$cluster, dist(data_limited))
p_s <- fviz_silhouette(sil, print.summary = print.summary)
p_c <- fviz_cluster(km.res, geom = "point", data = data_limited) + ggtitle(paste('k =', centers))
data_with_clusters <- bind_cols(data_limited, cluster = as.factor(as.numeric(km.res$cluster)))
return(list(
km = km.res,
cluster = km.res$cluster,
data = data_with_clusters,
p_s = p_s,
p_c = p_c,
p_sc = ggarrange(p_s, p_c, ncol = 2)
))
}
method_study <- function(filter_data, names, centers = 7) {
m_num <- method_number(filter_data, names, centers)
m_calc <- method_calculate(filter_data, names, centers)
return(list(
plot = ggarrange(m_num$p_elbow, m_num$p_silhouette, m_num$p_gap_statistic, m_num$p_NbClust,
m_calc$p_s, m_calc$p_c,
nrow = 2, ncol = 3),
num = m_num,
calc = m_calc
))
}
## Columns names ----
names_gdp <- c('penergy', 'pmining', 'pmetal', 'pmanufac', 'pbuilding', 'pagric', 'ptextile', 'ppharmac', 'pdurab', 'pinterind', 'pinindother', 'pothwhole', 'pretail_food', 'pretail_nonf', 'pretail_other')
names_people <- c('pmale', 'pob', 'pforeign', 'pobgrowth', 'unemprate')
names_ind <- c('ind_wholesale', 'ind_retail', 'ind_rest', 'ind_turis', 'ind_actindex')
names_rich <- c('adsl', 'pcars', 'pbanks')
names_all <- c(names_gdp, names_people, names_ind, names_rich)
names_max_sd <- c('pob', 'ind_actindex', 'pbanks', 'ptextile', 'pothwhole', 'pretail_other')
##Pre process ----
# Barcelona and Madrid are outliers
model_ <- preProcess(data, method = c("center", "scale"))
data_center <- predict(model_, data) %>%
filter(!provincia %in% c('Madrid', 'Barcelon')) %>%
select(-provincia)
# Trial and error
method_study(data_center, names_all, 5)$plot #0.14
method_study(data_center, names_gdp, 3)$plot #0.21
method_study(data_center, names_people, 4)$plot #0.36
method_study(data_center, names_ind, 3)$plot #0.60
method_study(data_center, names_rich, 7)$plot #0.33
method_study(data_center, names_all, 2)$plot #0.27
method_study(data_center, names_max_sd, 6)$plot #0.26
## Using RF for variable importance ----
# Recommendation by Jesús
num_clusters <- 5
t <- method_calculate(data_center, names_all, num_clusters)
# Variable importance to explain the kmeans
set.seed(42)
model <- caret::train(cluster ~ ., data = t$data, method = "rf", tuneLength = 40)
confusionMatrix(data = predict(model, t$data), reference = t$data$cluster)
names_sorted <- varImp(model) %>%
.$importance %>%
rownames_to_column(var = 'name') %>%
arrange(desc(Overall)) %>%
top_n(5, Overall)
method_study(data_center, names_sorted$name, 3)$plot #10 0.26
method_study(data_center, names_sorted$name, 3)$plot #15 0.20
method_study(data_center, names_sorted$name, 3)$plot #5 0.27
## Using PCA for variable importance ----
# https://stackoverflow.com/questions/26529659/variable-selection-for-k-means-clustering
#Variable selection
model_ <- preProcess(data_center, method = c("pca"), thresh = 1)
names_sorted <- apply(model_$rotation, 1, function(x) sum(x))
names_sorted <- data.frame(names = names(names_sorted), value = as.numeric(names_sorted)) %>%
dplyr::arrange(desc(value))
# Explained variance
# https://stats.stackexchange.com/questions/254592/calculating-pca-variance-explained/254598
model_ <- prcomp(data_center[, names_all], center = TRUE, scale. = TRUE)
names_sorted <- model_$rotation[, 'PC1']
names_sorted <- data.frame(names = names(names_sorted), value = as.numeric(abs(names_sorted))) %>%
dplyr::arrange(desc(value))
plot_cluster_selection_methods(data_center, names_sorted$names[1:13], 9) #0.86
plot_cluster_selection_methods(data_center, names_all, 9) #0.86
plot_sillhouette(data_center, names_sorted$names[1:13], 4)
## Using the centers ----
s <- plot_sillhouette(data_center, names_all, 5)
s1 <- apply(s$km$centers, 2, function(x)sum(abs(x)))
names_sorted <- data.frame(names = names(s1), value = as.numeric(s1)) %>%
dplyr::arrange(desc(value))
plot_cluster_selection_methods(data_center, names_sorted$names[1:13], 9) #0.86
plot_sillhouette(data_center, names_sorted$names[1:13], 9)