Arianto, Ahmad Dwi (2017) Klasifikasi Data Keuangan Sektor Publik Untuk Penentuan Sampel Pemeriksaan Menggunakan K-Nearest Neighbors. Masters thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
2215206715-Master_Theses.pdf - Published Version Download (2MB) | Preview |
Abstract
Penentuan sampel pemeriksaan internal yang dilakukan oleh Inspektorat
Utama BPK RI seringkali berdasarkan “kebiasaan” semata. Metode penambangan
data bisa menjadi salah satu cara alternatif untuk mewariskan “kebiasaan” tersebut
pada generasi penerus. Penelitian deteksi opini dan deteksi pemalsuan/kecurangan
laporan keuangan menggunakan teknik penambangan data (data mining) sudah
sangat lazim di sektor swasta, namun tidak di sektor publik. Orientasi pada laba
menjadi pembeda utama yang bermuara pada karakter laporan keuangan kedua
sektor tersebut. Klasifikasi dilakukan melalui dua tahap, tahap pertama klasifikasi
berdasarkan 2 kelas, dan tahap kedua klasifikasi berdasarkan 4 kelas. Validasi
silang 10 lipatan (10-folds cross validation) yang dipadukan dengan berbagai
skema normalisasi data, fitur, jarak, Nilai-K, dan uji lanjut akan digunakan untuk
membangun model terbaik. Hasil penelitian menunjukkan bahwa area di bawah
kurva/Area Under Curve (AUC) tertinggi untuk klasifikasi KNN 2 kelas adalah
66,30%, sedangkan AUC untuk 4 kelas adalah 61,56% yang termasuk klasifikasi
dengan kinerja buruk. Penelitian lebih lanjut diperlukan untuk meningkatkan
kinerja model prediksi opini dari penelitian ini.
=================================================================
The determination of the internal audit sample conducted by the Main
Inspectorate of BPK RI is often based on the mere “habits”. Data mining can be one
of the alternative ways to pass “the habits” on the next generation. Research on
opinion detection and falsified/fraudulent financial statement detection using data
mining techniques is very common in the private sector, but not in the public sector.
Profit orientation was a key differentiator that led to the different characteristics of
financial statements in the both sector. The classification was done through two
stages, firstly classification based on 2 classes, and secondly classification based on
4 classes. Ten folds cross-validation combined with various data normalization
schemes, features schemes, distances schemes, K-values schemes, and posthoc test
schemes were used to build the best model. The results showed that The highest
Area Under Curve (AUC) for classification KNN 2 class was 66.30%, while the
highest AUC for 4 classes was 61.56% which fell under poor classification. Further
research is needed to improve the performance of classification models from this
study.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | sampel, penambangan data, opini, laporan keuangan sektor publik,KNN, sample, data mining, opinion, public sector financial statement |
| Subjects: | H Social Sciences > HJ Public Finance T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.546 Computer algorithms |
| Divisions: | Faculty of Industrial Technology > Electrical Engineering > 20101-(S2) Master Thesis |
| Depositing User: | - Ahmad Dwi Arianto |
| Date Deposited: | 14 Aug 2017 01:53 |
| Last Modified: | 14 Aug 2017 01:53 |
| URI: | http://repository.its.ac.id/id/eprint/41874 |
Actions (login required)
 |
View Item |