Implementasi Metode Bagging Nearest Neighbor Support Vector Machine Untuk Prediksi Kebangkrutan Penyusun: M. Ulin Nuha – 5108100164 Dosen Pembimbing: Isye Arieshanti, S.Kom., M.Phil Yudhi Purwananto, S.Kom., M.Kom. PRESENTASI TUGAS AKHIR – KI091391 (Keyword: Prediksi kebangkrutan, BNNSVM, Bootstrap aggregating, K-nearest neighbor, Support Vector Machine)
60
Embed
Implementasi Bagging Nearest Neighbor Support Vector ...digilib.its.ac.id/public/ITS-paper-25817-5108100164-Presentation.pdfPRESENTASI TUGAS AKHIR – KI091391 (Keyword: Prediksi kebangkrutan,
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
SVM Training SVM Training SVM Training SVM Training
SVM Model 1 SVM Model 2 SVM Model 9 SVM Model
10 …
Proses: Uji
SVM testing
Bagging
Data Uji
SVM Model 1 SVM Model 2 SVM Model 9 SVM Model
10 …
Proses: Uji
Bagging SVM
testing
Data Uji
SVM Model 1 SVM Model 2 SVM Model 9 SVM Model
10 …
Proses: Uji
Bagging SVM
testing
SVM Model 1 SVM Model 2 SVM Model 9 SVM Model
10 …
SVM Testing SVM Testing SVM Testing SVM Testing
Prediksi 1 Prediksi 2 Prediksi 9 Prediksi 10 …
Data Uji
SVM testing
Proses: Uji
SVM Model 1 SVM Model 2 SVM Model 9 SVM Model
10 …
SVM Testing SVM Testing SVM Testing SVM Testing
Prediksi 1 Prediksi 2 Prediksi 9 Prediksi 10 …
Data Uji
Bagging
SVM testing
Proses: Uji
Bagging
Prediksi 1 Prediksi 2 Prediksi 9 Prediksi 10 …
Bagging (voting)
Prediksi
Akhir
Antarmuka
Studi Literatur
Desain dan Implementasi
Nilai k (KNN)
Nilai cost & Jenis Kernel (SVM)
Perbandingan dengan metode lain
Uji Coba
Uji coba dengan nilai k berbeda
(Wieslaw) k Akurasi Presisi
Sensiti-
vity Specifi-
city
1 65.67 67.88 69.76 61.67
2 68.92 71.39 70.61 67.18
3 68.92 72.41 69.85 69.78
4 70.58 72.87 71.99 68.6
5 70.83 73.82 71.03 70.25
6 69.42 72.78 69.06 69.48
7 70.5 75.3 68.76 72.5
8 70.58 73.55 70.41 70.91
9 70.83 74.39 70.55 72.58
10 71.58 73.99 71.86 71.08
60
65
70
75
80
85
90
1 2 3 4 5 6 7 8 9 10
Akurasi Presisi
Sensitivity Specificity
Uji coba dengan nilai k berbeda
(Australian) k Akurasi Presisi
Sensiti
-vity
Specifi
-city
1 84.49 83.91 81.17 87.22
2 84.9 83.94 82.35 87.19
3 84.93 83.78 82.76 87.04
4 85.74 84.81 83.02 87.84
5 85.3 83.19 84.54 86.28
6 84.43 83.53 81.88 86.51
7 85.07 83.08 83.94 86.16
8 86.23 85.92 83.06 88.95
9 85.77 84.96 82.95 88.1
10 85.22 84.68 82.34 87.3
60
65
70
75
80
85
90
1 2 3 4 5 6 7 8 9 10
Akurasi Presisi
Sensitivity Specificity
Uji coba dengan nilai cost SVM
berbeda (Wieslaw)
Cost Akurasi Presisi Sensiti-
vity Specifi-
city
0.01 67.58 68.22 75.51 59.25
0.1 70.17 72.66 72.12 66.93
1 71.08 75.12 71.61 72.17
10 70.33 72.58 70.85 69.47
100 71 75.25 70.03 72.87
55
60
65
70
75
80
85
90
95
0.01 0.1 1 10 100
Akurasi Presisi
Sensitivity Specificity
Uji coba dengan nilai cost SVM
berbeda (Australian)
Cost Akurasi Presisi Sensiti-
vity Specifi-
city
0.01 83.51 86.86 74.5 90.7
0.1 85.54 82.51 85.55 85.44
1 84.64 83.64 82.09 86.56
10 80.72 81.68 73.79 86.27
100 75.8 78.57 65.46 83.85
55
60
65
70
75
80
85
90
95
0.01 0.1 1 10 100
Akurasi Presisi
Sensitivity Specificity
Uji coba dengan kernel RBF dan nilai
gamma berbeda (Wieslaw)
Gamma Akurasi Presisi Sensiti-
vity Specifi-
city
0.0001 54.17 58.36 73.19 35.07
0.001 58.75 60.59 65.61 50.55
0.01 56.83 57.28 74.14 37.45
0.1 54.5 54.36 89.48 14.33
1 53.25 53.3 99.87 0
10 52.67 51 96 4 0
20
40
60
80
100
Akurasi Presisi
Sensitivity Specificity
Uji coba dengan kernel RBF dan nilai
gamma berbeda (Australian)
Gamma Akurasi Presisi Sensiti-
vity Specifi-
city
0.0001 67.88 68.43 52.17 80.79
0.001 68.81 64.98 64.13 72.45
0.01 56.06 51.05 14.67 89.41
0.1 55.04 16 0.4 98.89
1 55.51 0 0 100
10 55.51 0 0 100 0
20
40
60
80
100
Akurasi Presisi
Sensitivity Specificity
Uji coba dengan kernel Polynomial
dan nilai degree berbeda (Wieslaw)
Degree Akurasi Presisi Sensiti-
vity Specifi-
city
1 65.25 66.92 71.77 58.96
2 71.33 72.34 74.22 67.8
3 69.67 72.22 69.36 69.95
4 68.08 71.42 68.64 68.47
5 70.08 72.25 71.11 69.32
50
60
70
80
90
100
1 2 3 4 5
Akurasi Presisi
Sensitivity Specificity
Uji coba dengan kernel Polynomial
dan nilai degree berbeda (Australian)
Degree Akurasi Presisi Sensiti-
vity Specifi-
city
1 79.83 89.47 62.87 93.85
2 80.32 85.79 67.64 90.48
3 72.35 75.74 62.65 79.87
4 60.09 64.66 62.55 57.88
5 57.97 64.9 52.6 62.46
50
60
70
80
90
100
1 2 3 4 5
Akurasi Presisi
Sensitivity Specificity
Uji coba perbandingan dengan
metode klasifikasi lain (Wieslaw)
Metode Akurasi Presisi Sensiti-
vity
Spesifi-
city
KNN 75 76.22 78.04 73.03
ANN 70 70 70 70
SVM 70.42 74.29 69.91 74.82
BLR 87.54 90.68 86.42 11.07
BNN-
SVM 71.58 73.99 71.86 71.08
65
70
75
80
85
90
95
100
Akurasi Presisi
Sensitivity Spesificity
KNN = K-Nearest Neighbor
ANN = Artificial Neural Network
SVM = Support Vector Machine
BLR = Binary Logistic Regression
BNNSVM = Bagging Nearest Neighbor
Support Vector Machine
Uji coba perbandingan dengan
metode klasifikasi lain (Australian)
Metode Akurasi Presisi Sensiti-
vity
Spesifi-
city
KNN 83.19 80.9 80.5 85.2
ANN 83.48 83.5 85.11 87.94
SVM 84.35 77.03 93.44 77.92
BLR 80.83 81.73 75.89 14.84
BNN-
SVM 86.23 85.92 83.06 88.95
65
70
75
80
85
90
95
100
Akurasi Presisi
Sensitivity Spesificity
KNN = K-Nearest Neighbor
ANN = Artificial Neural Network
SVM = Support Vector Machine
BLR = Binary Logistic Regression
BNNSVM = Bagging Nearest Neighbor
Support Vector Machine
Latar Belakang
Tujuan
Permasalahan
Pengembangan Perangkat Lunak
Daftar Pustaka
Kesimpulan
BNNSVM Prediksi
Kebangkrutan
Dataset Akurasi Presisi Sensiti-
vity
Specifi-
city
Wieslaw 71.58 % 73.99 % 71.86 % 71.08 %
Australian
credit
approval
86.23 % 85.92 % 83.06 % 88.95 %
Hasil
Latar Belakang
Tujuan
Permasalahan
Pengembangan Perangkat Lunak
Kesimpulan
Daftar Pustaka
Daftar Pustaka
Li, H., & Sun, J. (2011). Forecasting Business Failure: The Use of Nearest-Neighbour Support Vectors and Correcting Imbalanced Samples - Evidence from Chinese Hotel Industry. Tourism Management , XXXIII (3), 622-634.
Frank, A., & Asuncion, A. (2010). UCI Machine Learning Repository. University of California, Irvine, School of Information and Computer Sciences. Diambil kembali dari http://archive.ics.uci.edu/ml
Wieslaw, P. (2004). Application of Discrete Predicting Structures in An Early Warning Expert System for Financial Distress. Tourism Management.
Tan, P. N., Steinbach, M., & Kumar, V. (2006). Introduction to Data Mining (4th ed.). Boston: Pearson Addison Wesley.