Klasifikasi

Campus Recruitment Academic and Employability Factors influencing placement source data : https://www.kaggle.com/benroshan/factors-affecting-campus-placement

Suasana Kampus

Deskripsi Data

Variabel	Keterangan
ssc_p	Secondary Education percentage- 10th Grade
ssc_b	Board of Education- Central/ Others
hsc_p	Higher Secondary Education percentage- 12th Grade
hsc_b	Board of Education- Central/ Others
hsc_s	Specialization in Higher Secondary Education
degree_p	Degree Percentage
degree_t	Under Graduation(Degree type)- Field of degree education
workex	Work Experience
specialisation	Post Graduation(MBA)- Specialization
status	Status of placement- Placed/Not placed
salary	Salary offered by corporate to candidates
mba_p	MBA percentage

Dataset

Ringkasan Data

summary(placement_data)

     sl_no          gender              ssc_p          ssc_b          
 Min.   :  1.0   Length:215         Min.   :40.89   Length:215        
 1st Qu.: 54.5   Class :character   1st Qu.:60.60   Class :character  
 Median :108.0   Mode  :character   Median :67.00   Mode  :character  
 Mean   :108.0                      Mean   :67.30                     
 3rd Qu.:161.5                      3rd Qu.:75.70                     
 Max.   :215.0                      Max.   :89.40                     
                                                                      
     hsc_p          hsc_b              hsc_s              degree_p    
 Min.   :37.00   Length:215         Length:215         Min.   :50.00  
 1st Qu.:60.90   Class :character   Class :character   1st Qu.:61.00  
 Median :65.00   Mode  :character   Mode  :character   Median :66.00  
 Mean   :66.33                                         Mean   :66.37  
 3rd Qu.:73.00                                         3rd Qu.:72.00  
 Max.   :97.70                                         Max.   :91.00  
                                                                      
   degree_t         workex       etest_p     specialisation         mba_p      
 Length:215         No :141   Min.   :50.0   Length:215         Min.   :51.21  
 Class :character   Yes: 74   1st Qu.:60.0   Class :character   1st Qu.:57.95  
 Mode  :character             Median :71.0   Mode  :character   Median :62.00  
                              Mean   :72.1                      Mean   :62.28  
                              3rd Qu.:83.5                      3rd Qu.:66.25  
                              Max.   :98.0                      Max.   :77.89  
                                                                               
    status              salary      
 Length:215         Min.   :200000  
 Class :character   1st Qu.:240000  
 Mode  :character   Median :265000  
                    Mean   :288655  
                    3rd Qu.:300000  
                    Max.   :940000  
                    NA's   :67      

introduce(placement_data)

plot_intro(placement_data)

profile_missing(placement_data)

plot_missing(placement_data, title = "Variabel Data Hilang")

plot_density(placement_data, geom_density_args = list("fill" = "purple", "alpha" = 0.6), ncol = 2L, ggtheme=theme_gray(), title= "Distribusi Peluang Masing-masing Variabel")

Cek karakteristik data

placement_data$workex=as.factor(placement_data$workex)
placement_data$gender=as.factor(placement_data$gender)
str(placement_data)

'data.frame':   215 obs. of  15 variables:
 $ sl_no         : int  1 2 3 4 5 6 7 8 9 10 ...
 $ gender        : Factor w/ 2 levels "F","M": 2 2 2 2 2 2 1 2 2 2 ...
 $ ssc_p         : num  67 79.3 65 56 85.8 ...
 $ ssc_b         : chr  "Others" "Central" "Central" "Central" ...
 $ hsc_p         : num  91 78.3 68 52 73.6 ...
 $ hsc_b         : chr  "Others" "Others" "Central" "Central" ...
 $ hsc_s         : chr  "Commerce" "Science" "Arts" "Science" ...
 $ degree_p      : num  58 77.5 64 52 73.3 ...
 $ degree_t      : chr  "Sci&Tech" "Sci&Tech" "Comm&Mgmt" "Sci&Tech" ...
 $ workex        : Factor w/ 2 levels "No","Yes": 1 2 1 1 1 2 1 2 1 1 ...
 $ etest_p       : num  55 86.5 75 66 96.8 ...
 $ specialisation: chr  "Mkt&HR" "Mkt&Fin" "Mkt&Fin" "Mkt&HR" ...
 $ mba_p         : num  58.8 66.3 57.8 59.4 55.5 ...
 $ status        : chr  "Placed" "Placed" "Placed" "Not Placed" ...
 $ salary        : int  270000 200000 250000 NA 425000 NA NA 252000 231000 NA ...

Pada data di atas tidak memiliki data yang kosong akan tetapi data yang tidak diketahui berjumlah 67. Data yang tidak diketahui tersebut terdapat dalam variabel ‘salary’. Oleh karenanya 67 data tersebut tidak memiliki status.

placement_data %>% group_by(status) %>% count()

Analisis Deskriptif

Korelasi

placement_data_corr = select(placement_data, ssc_p, hsc_p, degree_p, etest_p, mba_p)
corr <- round(cor(placement_data_corr),2)
ggcorrplot::ggcorrplot(corr, method = "square" , ggtheme = theme_update(), lab = TRUE,
           lab_size = 5, type = "lower", colors = c("blue", "white", "purple"),
           digits = 2)

Persentase Status

Gender

placement_data %>%
  count(gender, status) %>%
  group_by(gender) %>%
  mutate(n = n/sum(n) * 100) %>%
  ggplot() + aes(gender, n, fill = status, label = paste0(round(n, 2), "%")) + 
  geom_col() +
  geom_text(position=position_stack(0.5))+labs(x="Gender", y="Banyaknya gender", title = "Persentase Status Terhadap Gender")

ggplot(placement_data, aes(x = factor(gender), fill = gender)) +
  geom_bar() +
  geom_text(aes(label = ..count..), stat = "count", vjust = 1.5, colour = "white")+ labs( x = "Gender", y = "Jumlah")+ggtitle("Rasio Variabel gender ")

hsc_b

placement_data %>%
  count(hsc_b, status) %>%
  group_by(hsc_b) %>%
  mutate(n = n/sum(n) * 100) %>%
  ggplot() + aes(hsc_b, n, fill = status, label = paste0(round(n, 2), "%")) + 
  geom_col() +
  geom_text(position=position_stack(0.5))+labs(x="Board of Education", y="Banyaknya Board of Education", title = "Persentase Status Terhadap Board of Education ")

ggplot(placement_data, aes(x = factor(hsc_b), fill = hsc_b)) +
  geom_bar() +
  geom_text(aes(label = ..count..), stat = "count", vjust = 1.5, colour = "white")+ labs( x = "Board of Education", y = "Jumlah")+ggtitle("Rasio Variabel Board of Education")

hsc_s

placement_data %>%
  count(hsc_s, status) %>%
  group_by(hsc_s) %>%
  mutate(n = n/sum(n) * 100) %>%
  ggplot() + aes(hsc_s, n, fill = status, label = paste0(round(n, 2), "%")) + 
  geom_col() +
  geom_text(position=position_stack(0.5))+labs(x="Specialization in Higher Secondary Education ", y="Banyaknya Specialization in Higher Secondary Education ", title = "Persentase Status terhadap Specialization in Higher Secondary Education")

ggplot(placement_data, aes(x = factor(hsc_s), fill = hsc_s)) +
  geom_bar() +
  geom_text(aes(label = ..count..), stat = "count", vjust = 1.5, colour = "white")+ labs( x = "Specialization in Higher Secondary Education", y = "Jumlah")+ggtitle("Rasio Variabel Specialization in Higher Secondary Education")

degree_t

placement_data %>%
  count(degree_t, status) %>%
  group_by(degree_t) %>%
  mutate(n = n/sum(n) * 100) %>%
  ggplot() + aes(degree_t, n, fill = status, label = paste0(round(n, 2), "%")) + 
  geom_col() +
  geom_text(position=position_stack(0.5))+labs(x="Under Graduation ", y="Banyaknya Under Graduation", title = "Persentase Status terhadap Under Graduation")

ggplot(placement_data, aes(x = factor(degree_t), fill = degree_t)) +
  geom_bar() +
  geom_text(aes(label = ..count..), stat = "count", vjust = 1.5, colour = "white")+ labs( x = "Under Graduation", y = "Jumlah")+ggtitle("Rasio Variabel Under Graduation")

workex

placement_data %>%
  count(workex, status) %>%
  group_by(workex) %>%
  mutate(n = n/sum(n) * 100) %>%
  ggplot() + aes(workex, n, fill = status, label = paste0(round(n, 2), "%")) + 
  geom_col() +
  geom_text(position=position_stack(0.5))+labs(x="Work Experience ", y="Banyaknya Work Experience", title = "Persentase Status terhadap Work Experience")

ggplot(placement_data, aes(x = factor(workex), fill = workex)) +
  geom_bar() +
  geom_text(aes(label = ..count..), stat = "count", vjust = 1.5, colour = "white")+ labs( x = "Work Experience", y = "Jumlah")+ggtitle("Rasio Variabel Work Experience")

specialisation

placement_data %>%
  count(specialisation, status) %>%
  group_by(specialisation) %>%
  mutate(n = n/sum(n) * 100) %>%
  ggplot() + aes(specialisation, n, fill = status, label = paste0(round(n, 2), "%")) + 
  geom_col() +
  geom_text(position=position_stack(0.5))+labs(x="Post Graduation(MBA) ", y="Banyaknya Post Graduation(MBA)", title = "Persentase Status terhadap Post Graduation(MBA)")

ggplot(placement_data, aes(x = factor(specialisation), fill = specialisation)) +
  geom_bar() +
  geom_text(aes(label = ..count..), stat = "count", vjust = 1.5, colour = "white")+ labs( x = "Post Graduation(MBA)", y = "Jumlah")+ggtitle("Rasio variabel Post Graduation(MBA)")

Rataaan Pendidikan

placement_data %>%
  mutate(rataan_total = (ssc_p+hsc_p+degree_p+mba_p)/4)%>%
  ggplot(aes(rataan_total, fill = status))+
  geom_histogram(binwidth = 5, col="black")+
  # scale_fill_manual(values = c("#DC3220", "#40B0A6"))+
  # scale_colour_manual(values = c("#DC3220", "#40B0A6"))+
  labs(x = "Rata-rata nilai 4 sistem pendidikan",
       y = "Banyaknya murid",
       fill = "Status",
       title = "Rata-rata hasil dari 'ssc_p, hsc_p, degree_p, dan mba_p'")

Analisis Klasifikasi

Model C5.0

Karena data yang tidak diketahui berjumlah 67 dari total data sebanyak 215, maka data yang bisa diolah berjumlah 148.

Rasio data latihan dengan data prediksi adalah 80:20 data latihan = 173 data prediksi = 42

set.seed(100)
placement_data_olah <- placement_data %>%
  select(-salary)%>%
  mutate(status = as.factor(make.names(status)))


split <- createDataPartition(placement_data_olah$status,
                             p =0.8, 
                             list = FALSE)
data_latihan = placement_data_olah[split, ]
data_prediksi = placement_data_olah[-split,]

class_data_latihan <- data_latihan$status
# data_latihan =as.factor(data_latihan)

data_latihan <- data_latihan%>%select(-sl_no,-status)
data_prediksi <- data_prediksi%>%select(-sl_no)
placement_data_c50 <- C5.0(data_latihan, class_data_latihan)
summary(placement_data_c50)

Call:
C5.0.default(x = data_latihan, y = class_data_latihan)


C5.0 [Release 2.07 GPL Edition]     Sun Jul 25 09:40:37 2021
-------------------------------

Class specified by attribute `outcome'

Read 173 cases (13 attributes) from undefined.data

Decision tree:

ssc_p <= 56.28: Not.Placed (27/1)
ssc_p > 56.28:
:...workex = Yes: Placed (52/1)
    workex = No:
    :...hsc_p <= 52: Not.Placed (7)
        hsc_p > 52:
        :...degree_p > 65: Placed (48/4)
            degree_p <= 65:
            :...hsc_p > 70.4: Placed (9)
                hsc_p <= 70.4:
                :...mba_p <= 57.99: Placed (12/2)
                    mba_p > 57.99:
                    :...ssc_p <= 70.5: Not.Placed (14/1)
                        ssc_p > 70.5: Placed (4/1)


Evaluation on training data (173 cases):

        Decision Tree   
      ----------------  
      Size      Errors  

         8   10( 5.8%)   <<


       (a)   (b)    <-classified as
      ----  ----
        46     8    (a): class Not.Placed
         2   117    (b): class Placed


    Attribute usage:

    100.00% ssc_p
     84.39% workex
     54.34% hsc_p
     50.29% degree_p
     17.34% mba_p


Time: 0.0 secs

plot(placement_data_c50)

Model C5.0 Boost

dilakukan pengulangan sebanyak 10 kali, hasilnya data menjadi lebih akurat

placement_data_boost <- C5.0(data_latihan, class_data_latihan, trials = 10)
summary(placement_data_boost)

Call:
C5.0.default(x = data_latihan, y = class_data_latihan, trials = 10)


C5.0 [Release 2.07 GPL Edition]     Sun Jul 25 09:40:51 2021
-------------------------------

Class specified by attribute `outcome'

Read 173 cases (13 attributes) from undefined.data

-----  Trial 0:  -----

Decision tree:

ssc_p <= 56.28: Not.Placed (27/1)
ssc_p > 56.28:
:...workex = Yes: Placed (52/1)
    workex = No:
    :...hsc_p <= 52: Not.Placed (7)
        hsc_p > 52:
        :...degree_p > 65: Placed (48/4)
            degree_p <= 65:
            :...hsc_p > 70.4: Placed (9)
                hsc_p <= 70.4:
                :...mba_p <= 57.99: Placed (12/2)
                    mba_p > 57.99:
                    :...ssc_p <= 70.5: Not.Placed (14/1)
                        ssc_p > 70.5: Placed (4/1)

-----  Trial 1:  -----

Decision tree:

ssc_p > 77.8: Placed (24.5)
ssc_p <= 77.8:
:...ssc_p <= 52.58: Not.Placed (13.8)
    ssc_p > 52.58:
    :...workex = Yes: Placed (33.9/5.6)
        workex = No:
        :...ssc_b = Others: Placed (29.3/8.4)
            ssc_b = Central:
            :...ssc_p <= 52.6: Placed (4.8)
                ssc_p > 52.6:
                :...degree_t in {Sci&Tech,Others}: Not.Placed (27/2.3)
                    degree_t = Comm&Mgmt:
                    :...ssc_p <= 63.4: Not.Placed (26.7/6.9)
                        ssc_p > 63.4: Placed (13/1.5)

-----  Trial 2:  -----

Decision tree:

ssc_p > 77.8: Placed (19.4)
ssc_p <= 77.8:
:...ssc_p <= 52.58: Not.Placed (10.9)
    ssc_p > 52.58:
    :...hsc_p <= 64.2:
        :...hsc_p > 63: Not.Placed (19.3)
        :   hsc_p <= 63:
        :   :...hsc_p <= 54: Not.Placed (15.1/0.6)
        :       hsc_p > 54: Placed (36.6/13.3)
        hsc_p > 64.2:
        :...hsc_s in {Science,Arts}: Placed (21.9)
            hsc_s = Commerce:
            :...ssc_p > 67: Placed (11.5)
                ssc_p <= 67:
                :...degree_p <= 64.5: Not.Placed (13.8/1.8)
                    degree_p > 64.5: Placed (24.5/8.9)

-----  Trial 3:  -----

Decision tree:

ssc_p > 77.8: Placed (15.3)
ssc_p <= 77.8:
:...workex = Yes:
    :...ssc_p <= 56.6: Not.Placed (4.9)
    :   ssc_p > 56.6: Placed (26.6/4.7)
    workex = No:
    :...mba_p > 67.2: Not.Placed (23.9/1)
        mba_p <= 67.2:
        :...hsc_p > 64.2: Placed (48.2/15.8)
            hsc_p <= 64.2:
            :...degree_p <= 65: Not.Placed (43.9/5.4)
                degree_p > 65: Placed (10.3/3.5)

-----  Trial 4:  -----

Decision tree:

ssc_p > 77.8: Placed (12.3)
ssc_p <= 77.8:
:...ssc_p <= 56.28: Not.Placed (23.5/2.4)
    ssc_p > 56.28:
    :...hsc_p > 78.5: Placed (10.6)
        hsc_p <= 78.5:
        :...mba_p > 58.23: Not.Placed (92.2/27.9)
            mba_p <= 58.23:
            :...mba_p <= 52.21: Not.Placed (4.4)
                mba_p > 52.21: Placed (29.9/2.8)

-----  Trial 5:  -----

Decision tree:

workex = Yes: Placed (41.2/8.1)
workex = No:
:...hsc_p <= 52: Not.Placed (9.5)
    hsc_p > 52:
    :...ssc_b = Others: Placed (43.2/13.8)
        ssc_b = Central:
        :...mba_p > 66.06: Not.Placed (24.3/3.6)
            mba_p <= 66.06:
            :...gender = F: Placed (15.1/2.5)
                gender = M:
                :...degree_p <= 63.35: Not.Placed (16/1.5)
                    degree_p > 63.35: Placed (23.6/7.4)

-----  Trial 6:  -----

Decision tree:

ssc_p > 64:
:...hsc_p > 64: Placed (34.2/1.4)
:   hsc_p <= 64:
:   :...hsc_p <= 63: Placed (33.4/11)
:       hsc_p > 63: Not.Placed (11.5/0.3)
ssc_p <= 64:
:...hsc_p <= 59: Not.Placed (18.7)
    hsc_p > 59:
    :...degree_p > 73.43: Not.Placed (9.8)
        degree_p <= 73.43:
        :...degree_p <= 65: Not.Placed (49.2/14.3)
            degree_p > 65: Placed (16.3/2.6)

-----  Trial 7:  -----

Decision tree:

ssc_p <= 56.28: Not.Placed (25.6/1.9)
ssc_p > 56.28:
:...workex = Yes: Placed (35/4.7)
    workex = No:
    :...hsc_p > 70.2: Placed (21.9/2.5)
        hsc_p <= 70.2:
        :...mba_p <= 57.99: Placed (24.9/4.9)
            mba_p > 57.99: Not.Placed (65.7/17.1)

-----  Trial 8:  -----

Decision tree:

ssc_p <= 56.28: Not.Placed (22/3)
ssc_p > 56.28:
:...ssc_p > 77.8: Placed (13.4)
    ssc_p <= 77.8:
    :...mba_p > 68.81: Not.Placed (19.5/3.1)
        mba_p <= 68.81:
        :...workex = Yes: Placed (17.8)
            workex = No:
            :...hsc_p <= 54: Not.Placed (8)
                hsc_p > 54:
                :...hsc_p > 70.2: Placed (13.1)
                    hsc_p <= 70.2:
                    :...hsc_p > 69.4: Not.Placed (8.1/0.5)
                        hsc_p <= 69.4:
                        :...degree_t = Sci&Tech: Not.Placed (27.5/11.4)
                            degree_t in {Comm&Mgmt,Others}: Placed (43.5/8.1)

-----  Trial 9:  -----

Decision tree:

degree_p > 65:
:...ssc_b = Others: Placed (30.6/1)
:   ssc_b = Central:
:   :...workex = Yes: Placed (12.5)
:       workex = No:
:       :...etest_p <= 74.4: Placed (17.6/1.5)
:           etest_p > 74.4: Not.Placed (23.3/7.5)
degree_p <= 65:
:...ssc_p <= 56.6: Not.Placed (11.6)
    ssc_p > 56.6:
    :...hsc_p > 70.4: Placed (8.7)
        hsc_p <= 70.4:
        :...mba_p > 67.69: Not.Placed (10.3)
            mba_p <= 67.69:
            :...workex = Yes: Placed (7.8)
                workex = No:
                :...mba_p > 66.06: Placed (6.4)
                    mba_p <= 66.06:
                    :...ssc_p > 68: Placed (4.7)
                        ssc_p <= 68:
                        :...mba_p <= 57.99: Placed (16.6/6.3)
                            mba_p > 57.99: Not.Placed (23)


Evaluation on training data (173 cases):

Trial       Decision Tree   
-----     ----------------  
      Size      Errors  

   0         8   10( 5.8%)
   1         8   27(15.6%)
   2         9   16( 9.2%)
   3         7   20(11.6%)
   4         6   49(28.3%)
   5         7   27(15.6%)
   6         7   21(12.1%)
   7         5   27(15.6%)
   8         9   19(11.0%)
   9        12   14( 8.1%)
boost             0( 0.0%)   <<


       (a)   (b)    <-classified as
      ----  ----
        54          (a): class Not.Placed
             119    (b): class Placed


    Attribute usage:

    100.00% ssc_p
    100.00% hsc_p
    100.00% degree_p
    100.00% workex
     83.24% ssc_b
     83.24% mba_p
     39.88% degree_t
     38.15% hsc_s
     25.43% gender
     15.61% etest_p


Time: 0.0 secs

plot(placement_data_boost)

Model Trees

placement_data_tree = tree(status~., placement_data_olah[-1])
summary(placement_data_tree)

Classification tree:
tree(formula = status ~ ., data = placement_data_olah[-1])
Variables actually used in tree construction:
[1] "ssc_p"    "hsc_p"    "mba_p"    "workex"   "degree_p"
Number of terminal nodes:  13 
Residual mean deviance:  0.3372 = 68.12 / 202 
Misclassification error rate: 0.07442 = 16 / 215 

plot(placement_data_tree)
text(placement_data_tree, pretty = 0)

Prediksi Klasifikasi

Model C50

prediksi <- predict(placement_data_c50, data_prediksi)
# summary(prediksi)
cm_c50=confusionMatrix(prediksi, data_prediksi$status)
cm= draw_confusion_matrix(cm_c50)

Model Boost

prediksi_boost <- predict(placement_data_boost, data_prediksi)
# summary(prediksi_boost)

cm_boost=confusionMatrix(prediksi_boost, data_prediksi$status)
c= draw_confusion_matrix(cm_boost)

Model Tree

data_prediksi = placement_data_olah[-split,]
p=predict(placement_data_tree, data_prediksi, type = 'class')
cm_tree=confusionMatrix(p, data_prediksi$status, positive="Placed")
draw_confusion_matrix(cm_tree)

Hasil Prediksi

akurasi_c50 = as.data.frame(cm_c50$overall)[1,]
akurasi_boost = as.data.frame(cm_boost$overall)[1,]
akurasi_tree = as.data.frame(cm_tree$overall)[1,]

kappa_c50 = as.data.frame(cm_c50$overall)[2,]
kappa_boost = as.data.frame(cm_boost$overall)[2,]
kappa_tree = as.data.frame(cm_tree$overall)[2,]


nama = c("c50", "boost", "tree")
akurasi = c(akurasi_c50, akurasi_boost, akurasi_tree)
kappa = c(kappa_c50, kappa_boost, kappa_tree)
# hasil <- table(df$row_variable, df$column_variable)
data.frame(nama, akurasi, kappa)

Karena berdasarkan metode klasifikasi di atas penggunaan metode tree menunjukkan hasil terbaik, maka keputusan model yang dapat diambil adalah metode tree.

Sehingga, faktor yang mempengaruhi pertama kali untuk mendapatkan tempat adalah Persentase pendidikan menengah - kelas 10 (ssc_p)