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”

Uvoz podatkov

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

Urejanje podatkov

# 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.