Raziskovalno vprašanje: V koliko skupin lahko razdelimo enote opazovanja glede na izbrane spremenljivke - jabolka (velikost, teža, sladkost in hrustljavost) ter kakšna je kvaliteta le teh znotraj posamezne skupine.
Na to vprašanje sem se odločila odgovoriti z analizo razvrščanja po skupinah. Ta metoda mi bo pomagala razvrsiti enote po skupinah, glede na podane karaktaristike, ki sem si jih izbrala za analizo.
Podatki so bili pridobljeni iz spletne strani Kaggle.com “https://www.kaggle.com/datasets/nelgiriyewithana/apple-quality/data”
apples <- read.csv("apple_quality.csv")
head(apples)
## A_id Size Weight Sweetness Crunchiness Juiciness Ripeness
## 1 0 -3.9700485 -2.512336 5.3463296 -1.01200871 1.8449004 0.32983980
## 2 1 -1.1952172 -2.839257 3.6640588 1.58823231 0.8532858 0.86753008
## 3 2 -0.2920239 -1.351282 -1.7384292 -0.34261593 2.8386355 -0.03803333
## 4 3 -0.6571958 -2.271627 1.3248738 -0.09787472 3.6379705 -3.41376134
## 5 4 1.3642168 -1.296612 -0.3846582 -0.55300577 3.0308744 -1.30384943
## 6 5 -3.4253998 -1.409082 -1.9135112 -0.55577486 -3.8530715 1.91461592
## Acidity Quality
## 1 -0.4915905 good
## 2 -0.7228094 good
## 3 2.6216365 bad
## 4 0.7907232 good
## 5 0.5019840 good
## 6 -2.9815232 bad
Opis spremenljivk:
*Spremenljivke size, weight, sweetness, crunchiness, juiciness, ripeness in acidity so že standardizirane.
Velikost vzorca: 249 enot
# 7.1 clean the data
apples_clean <- subset(apples, select = -c(Juiciness, Ripeness, Acidity))
# 7.2 new dummy variable
apples_clean$quality_dummy <- ifelse(apples_clean$Quality == "good", 1, 0)
# 7.3 remove missing values
apples_clean <- na.omit(apples_clean)
summary(apples_clean)
## A_id Size Weight Sweetness
## Min. : 0 Min. :-4.9885 Min. :-6.2351 Min. :-5.0189
## 1st Qu.: 62 1st Qu.:-2.0021 1st Qu.:-2.0848 1st Qu.:-1.7384
## Median :124 Median :-0.5888 Median :-1.1220 Median :-0.6608
## Mean :124 Mean :-0.5667 Mean :-1.0334 Mean :-0.4509
## 3rd Qu.:186 3rd Qu.: 0.8079 3rd Qu.:-0.1284 3rd Qu.: 0.7949
## Max. :248 Max. : 4.4910 Max. : 5.1490 Max. : 5.3463
## Crunchiness Quality quality_dummy
## Min. :-3.2923 Length:249 Min. :0.0000
## 1st Qu.:-0.1953 Class :character 1st Qu.:0.0000
## Median : 0.9305 Mode :character Median :0.0000
## Mean : 0.9583 Mean :0.4819
## 3rd Qu.: 1.9325 3rd Qu.:1.0000
## Max. : 7.6199 Max. :1.0000
Opazimo, da povprečje ni 0, zato ponovno standardiziramo podatke.
std_apples <- apples_clean
std_apples$Size_z <- scale(std_apples$Size)
std_apples$Weight_z <- scale(std_apples$Weight)
std_apples$Sweetness_z <- scale(std_apples$Sweetness)
std_apples$Crunchiness_z <- scale(std_apples$Crunchiness)
summary(std_apples)
## A_id Size Weight Sweetness
## Min. : 0 Min. :-4.9885 Min. :-6.2351 Min. :-5.0189
## 1st Qu.: 62 1st Qu.:-2.0021 1st Qu.:-2.0848 1st Qu.:-1.7384
## Median :124 Median :-0.5888 Median :-1.1220 Median :-0.6608
## Mean :124 Mean :-0.5667 Mean :-1.0334 Mean :-0.4509
## 3rd Qu.:186 3rd Qu.: 0.8079 3rd Qu.:-0.1284 3rd Qu.: 0.7949
## Max. :248 Max. : 4.4910 Max. : 5.1490 Max. : 5.3463
## Crunchiness Quality quality_dummy Size_z.V1
## Min. :-3.2923 Length:249 Min. :0.0000 Min. :-2.3370055
## 1st Qu.:-0.1953 Class :character 1st Qu.:0.0000 1st Qu.:-0.7586627
## Median : 0.9305 Mode :character Median :0.0000 Median :-0.0116778
## Mean : 0.9583 Mean :0.4819 Mean : 0.0000000
## 3rd Qu.: 1.9325 3rd Qu.:1.0000 3rd Qu.: 0.7264842
## Max. : 7.6199 Max. :1.0000 Max. : 2.6730967
## Weight_z.V1 Sweetness_z.V1 Crunchiness_z.V1
## Min. :-3.498828 Min. :-2.3498018 Min. :-2.685489
## 1st Qu.:-0.707187 1st Qu.:-0.6623229 1st Qu.:-0.728845
## Median :-0.059607 Median :-0.1079942 Median :-0.017594
## Mean : 0.000000 Mean : 0.0000000 Mean : 0.000000
## 3rd Qu.: 0.608675 3rd Qu.: 0.6408292 3rd Qu.: 0.615447
## Max. : 4.158426 Max. : 2.9820994 Max. : 4.208633
Jabolka v vzorcu so podpovprečno velika,težka in sladka ter nadpovprečno hrustljava. Vidimo lahko, da je 48% jabolk dobre kvalitete.
library(fpc)
library(ggplot2)
library(factoextra)
## Welcome! Want to learn more? See two factoextra-related books at https://goo.gl/ve3WBa
distances <- get_dist(std_apples[c("Size_z", "Weight_z", "Sweetness_z", "Crunchiness_z")],
method = "euclidian")
distances2 <- distances*distances
fviz_dist(distances2)
Iz matrike razvidno bi lahko enote razporedili v 6 skupin
get_clust_tendency(std_apples[c("Size_z", "Weight_z", "Sweetness_z", "Crunchiness_z")],
n = nrow(std_apples) - 1,
graph = FALSE)
## $hopkins_stat
## [1] 0.7225424
##
## $plot
## NULL
Hopkinsova statistika je blizu 1, to pomeni, da so podatki primerni.
library(dplyr)
##
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
##
## filter, lag
## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, union
WARD <- std_apples[c("Size_z", "Weight_z", "Sweetness_z", "Crunchiness_z")] %>%
get_dist(method = "euclidian") %>%
hclust(method = "ward.D2")
WARD
##
## Call:
## hclust(d = ., method = "ward.D2")
##
## Cluster method : ward.D2
## Distance : euclidean
## Number of objects: 249
V 243 korakih so enote oblikovane v 6 skupin.
library(factoextra)
fviz_dend(WARD)
## Warning: The `<scale>` argument of `guides()` cannot be `FALSE`. Use "none" instead as
## of ggplot2 3.3.4.
## ℹ The deprecated feature was likely used in the factoextra package.
## Please report the issue at <https://github.com/kassambara/factoextra/issues>.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.
Razrez bi naredila pri 9, to bi imela 6 skupin.
std_apples$razvrstitevWARD =cutree (WARD,
k = 6)
head(std_apples)
## A_id Size Weight Sweetness Crunchiness Quality quality_dummy
## 1 0 -3.9700485 -2.512336 5.3463296 -1.01200871 good 1
## 2 1 -1.1952172 -2.839257 3.6640588 1.58823231 good 1
## 3 2 -0.2920239 -1.351282 -1.7384292 -0.34261593 bad 0
## 4 3 -0.6571958 -2.271627 1.3248738 -0.09787472 good 1
## 5 4 1.3642168 -1.296612 -0.3846582 -0.55300577 good 1
## 6 5 -3.4253998 -1.409082 -1.9135112 -0.55577486 bad 0
## Size_z Weight_z Sweetness_z Crunchiness_z razvrstitevWARD
## 1 -1.79874826 -0.9947933 2.98209939 -1.2448159 1
## 2 -0.33218549 -1.2146885 2.11673396 0.3979681 1
## 3 0.14517315 -0.2138371 -0.66232286 -0.8219059 2
## 4 -0.04782868 -0.8328854 0.91344987 -0.6672830 1
## 5 1.02053480 -0.1770645 0.03406119 -0.9548263 3
## 6 -1.51088875 -0.2527150 -0.75238561 -0.9565758 2
zac_voditelji <- aggregate(std_apples[, c("Size_z", "Weight_z", "Sweetness_z", "Crunchiness_z")],
by = list (std_apples$razvrstitevWARD),
FUN = mean)
head(zac_voditelji)
## Group.1 Size_z Weight_z Sweetness_z Crunchiness_z
## 1 1 0.002113889 -1.02001335 1.3518512 0.1571200
## 2 2 -0.384229130 -0.52767974 -0.8921624 -0.5215086
## 3 3 0.913716753 0.07085705 -0.1125733 -1.3652102
## 4 4 -1.181159324 0.65345492 0.5271927 -0.4316865
## 5 5 1.522170045 0.24942993 -1.0150718 0.4937618
## 6 6 0.112837968 0.22984775 -0.1115045 0.8718572
library(factoextra)
metodaVod <- hkmeans (std_apples[, c("Size_z", "Weight_z", "Sweetness_z", "Crunchiness_z")],
k = 6,
hc.metric = "euclidean",
hc.method = "ward.D2")
head(metodaVod)
## $cluster
## 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
## 4 1 2 1 3 2 3 4 4 2 2 5 5 1 5 1 2 1 6 6
## 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
## 2 5 4 2 5 2 6 5 6 6 2 4 3 1 1 2 6 5 1 1
## 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
## 5 1 1 6 6 6 6 6 4 6 4 2 5 5 3 5 3 1 6 3
## 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80
## 4 4 4 3 4 4 6 6 2 5 1 3 2 1 2 5 3 3 2 1
## 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100
## 3 1 3 1 5 2 5 3 2 5 6 2 5 1 2 5 2 5 6 4
## 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120
## 6 1 1 4 4 2 1 3 1 1 3 6 3 2 2 5 2 6 4 2
## 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140
## 4 6 4 5 5 4 4 4 4 1 4 6 4 5 2 3 2 1 6 6
## 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160
## 5 2 3 6 2 4 5 1 3 5 1 6 3 4 1 4 3 6 2 6
## 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180
## 2 5 1 5 6 6 1 5 5 4 1 2 6 2 1 4 1 4 2 1
## 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200
## 5 1 4 4 4 2 6 2 4 4 1 3 4 4 4 5 4 4 2 4
## 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220
## 2 4 1 2 1 5 4 5 1 2 6 2 3 6 1 5 6 3 5 1
## 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240
## 6 3 5 3 3 2 5 2 6 6 6 1 6 3 6 4 4 5 5 6
## 241 242 243 244 245 246 247 248 249
## 2 1 5 4 4 5 3 6 2
##
## $centers
## Size_z Weight_z Sweetness_z Crunchiness_z
## 1 0.04833115 -0.9084949 1.1036013 0.2828256
## 2 -0.34048346 -0.4631810 -0.9442028 -0.5108806
## 3 1.05597201 0.1621545 -0.2918755 -1.3191546
## 4 -1.26297903 0.4012258 0.7715501 -0.5149593
## 5 1.06616023 -0.1994924 -0.5530253 0.9487837
## 6 -0.06507878 1.0401632 -0.2217938 0.7776228
##
## $totss
## [1] 992
##
## $withinss
## [1] 74.58663 60.11274 74.02076 80.34033 79.54921 81.35145
##
## $tot.withinss
## [1] 449.9611
##
## $betweenss
## [1] 542.0389
Pod $centers vidimo končne voditelje.
fviz_cluster(metodaVod,
palette = "jama",
repel = FALSE,
ggtheme = theme_classic())
std_apples$razvrstitevVod <- metodaVod$cluster
head(std_apples[c("A_id", "razvrstitevWARD", "razvrstitevVod")])
## A_id razvrstitevWARD razvrstitevVod
## 1 0 1 4
## 2 1 1 1
## 3 2 2 2
## 4 3 1 1
## 5 4 3 3
## 6 5 2 2
Vidimo, da je bila prva enota prerazvrščena iz prve skupine v četrto.
table(std_apples$razvrstitevWARD, std_apples$razvrstitevVod)
##
## 1 2 3 4 5 6
## 1 28 0 0 4 3 0
## 2 3 37 0 0 0 0
## 3 1 1 26 0 0 0
## 4 0 0 0 43 0 6
## 5 0 0 3 0 19 4
## 6 11 7 0 0 20 33
Vidimo, da je veliko enot po razvrstitvi ostalo v osnovnih skupinah. Najbolj izstopajo enote, ki so bile iz 6 skupine prerazvrščene v 5 skupino (20enot) in 1 skupino (11 enot).
povprecja <- metodaVod$centers
head(povprecja)
## Size_z Weight_z Sweetness_z Crunchiness_z
## 1 0.04833115 -0.9084949 1.1036013 0.2828256
## 2 -0.34048346 -0.4631810 -0.9442028 -0.5108806
## 3 1.05597201 0.1621545 -0.2918755 -1.3191546
## 4 -1.26297903 0.4012258 0.7715501 -0.5149593
## 5 1.06616023 -0.1994924 -0.5530253 0.9487837
## 6 -0.06507878 1.0401632 -0.2217938 0.7776228
Položaji končnih voditeljev.
library(ggplot2)
library(tidyr)
slika <- as.data.frame(povprecja)
slika$A_id <- 1:nrow(slika)
slika <- pivot_longer(slika, cols = c(Size_z, Weight_z, Sweetness_z, Crunchiness_z))
slika$skupina <- factor(slika$A_id,
levels = c(1, 2, 3, 4, 5, 6),
labels = c("1", "2", "3", "4", "5", "6"))
slika$namefactor <- factor (slika$name,
levels = c("Size_z", "Weight_z", "Sweetness_z", "Crunchiness_z"),
labels = c("Size_z", "Weight_z", "Sweetness_z", "Crunchiness_z"))
ggplot(slika, aes(x = namefactor, y = value)) +
geom_hline(yintercept = 0) +
theme_bw() +
geom_point(aes(shape = skupina, col = skupina), size=3) +
geom_line(aes(group = A_id), linewidth = 1) +
ylab( "povprecje") +
xlab("razvrstitvene spremenljivke") +
ylim(-1.5, 1.5)
Opis značilnosti:
aggregate(std_apples$quality_dummy,
by = list(std_apples$razvrstitevVod),
FUN = mean)
## Group.1 x
## 1 1 0.65116279
## 2 2 0.02222222
## 3 3 0.82758621
## 4 4 0.40425532
## 5 5 0.57142857
## 6 6 0.55813953
Odgovor na raziskovalno vprašanje:
Opazovane enote smo s pomočjo dendrograma razdelili v 6 skupin glede na proučevane spremenljivke. V zadnjem izpisu lahko vidimo, da so bila v povprečju jabolka v 3 skupini okoli 82% dobra in 18% slaba. Najslabša jabolka pa so bila v 2 skupini, kjer je bilo v povprečju samo 2% dobrih jabolk in 98% slabih.