Evaluasi Performansi Klasifikasi Data Imbalanced dengan Robust Cost Sensitive Support Vector Machine

Fatika, Inggit (2025) Evaluasi Performansi Klasifikasi Data Imbalanced dengan Robust Cost Sensitive Support Vector Machine. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6003231026-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2027. Download (7MB) | Request a copy

Abstract

Data tidak seimbang menjadi tantangan dalam machine learning, terutama dalam hal klasifikasi di mana kelas minoritas sering kali memiliki dampak signifikan. Penelitian ini mengkaji Robust Cost-Sensitive Support Vector Machine (RCSSVM) sebagai solusi untuk menangani data tidak seimbang dan ketidakpastian dalam klasifikasi. RCSSVM mengintegrasikan pembelajaran cost-sensitive dengan optimasi robust, sehingga mampu menangani noise dan outlier pada data. Dengan memberikan penalti berbeda untuk kesalahan klasifikasi pada kelas minoritas dan mayoritas, RCSSVM memastikan keseimbangan perhatian terhadap kedua kelas. Penelitian ini menerapkan RCSSVM pada data financial distress dari 508 pemerintah daerah di Indonesia, yang memiliki rasio ketidakseimbangan 4:1 dengan jumlah daerah yang tidak mengalami financial distress jauh lebih banyak dibandingkan yang mengalami financial distress. Kinerja RCSSVM dievaluasi menggunakan berbagai kernel, yaitu linear, RBF, polinomial, dan sigmoid serta dibandingkan dengan metode lain seperti Robust SVM dan Cost-Sensitive SVM. Hasil menunjukkan bahwa RCSSVM secara konsisten unggul dibandingkan metode lainnya, dengan kernel linear memberikan performa terbaik berdasarkan G-mean (0,8268). Hal ini membuktikan kemampuan RCSSVM dalam menangani ketidakseimbangan data dan meningkatkan performa klasifikasi di bawah ketidakpastian ===================================================================================================================================
Imbalanced data remains a significant challenge in machine learning, particularly in classification tasks where the minority class holds critical importance. This study explores the Robust Cost-Sensitive Support Vector Machine (RCSSVM) as a solution to address imbalanced data and uncertainty in classification. RCSSVM integrates cost-sensitive learning with robust optimization, offering an effective mechanism for handling noisy data and outliers. By applying distinct penalties for classification errors in minority and majority classes, RCSSVM ensures a balanced focus on both. This research employs the RCSSVM method on financial distress data from 508 local governments in Indonesia, characterized by an imbalance ratio of 4:1, the number of non-financial distress region greater than those experience financial distress. The study evaluates the performance of RCSSVM using multiple kernels linear, RBF, polynomial, and sigmoid and compares its effectiveness against Robust SVM and Cost-Sensitive SVM. Results indicate that RCSSVM consistently outperforms other methods, with the linear kernel achieving the best outcomes G-Mean (0,8268). These findings demonstrate RCSSVM’s capability in addressing data imbalance and enhancing classification performance under uncertainty.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Data imbalanced, Klasifikasi, RCSSVM, Financial Distress Imbalanced Data, Classification, RCSSVM, Financial Distress
Subjects:	Q Science > QA Mathematics > QA353.K47 Kernel functions (analysis) Q Science > QA Mathematics > QA401 Mathematical models. Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science) Q Science > QA Mathematics > QA9.58 Algorithms
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis
Depositing User:	Fatika Inggit
Date Deposited:	06 Feb 2025 06:02
Last Modified:	06 Feb 2025 06:02
URI:	http://repository.its.ac.id/id/eprint/118369

Actions (login required)

View Item