Optimasi Metode Discriminatively Regularized Least Square Dengan Algoritma Genetika Dan Particle Swarm Optimizationuntvk Pengklasifikasian

Ririd, Ariadi Retno Tri Hayati (2010) Optimasi Metode Discriminatively Regularized Least Square Dengan Algoritma Genetika Dan Particle Swarm Optimizationuntvk Pengklasifikasian. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 5108201007-Master_thesis.pdf Restricted to Repository staff only Download (42MB)

Abstract

Metode regularization telah banyak diaplikasikan pada pengenalan pola yang bertujuan untuk menghasilkan klasifikasi data. Penelitian ini menggunakan metode Discriminatively Regularized Least Squares Classification (DRLSC) di mana pengklasifikasian datanya berdasarkan informasi discriminative yang dipengaruhi oleh jumlah data terdekat ($K$) berdasarkan metode K Nearest Neighbor dan regularization parameter ($\eta$). Penelitian ini mengoptimasi parameter $K$ dan $\eta$ secara otomatis berdasarkan error minimum pada metode pengklasifikasian DRLSC. Metode optimasi yang digunakan adalah penggabungan Algoritma Genetika dan Particle Swarm Optimization (GAPSO). Data yang digunakan untuk uji coba adalah database UCI, yaitu IRIS, WINE, dan LENSA. Berdasarkan hasil uji coba untuk mengoptimasi metode DRLSC, nilai fitness yang dihasilkan metode GAPSO lebih baik jika dibandingkan dengan Algoritma Genetika (GA) dan Particle Swarm Optimization (PSO) dengan selisih nilai fitness $7,3 \times 10^{-8}$ hingga $0,025$.
==================================================================================================================================
Regularization methods have been applied to pattern recognition which aims to classify data. This research uses the Discriminatively Regularized Least Squares Classification (DRLSC) method for classification, where the data classification is based on discriminative information influenced by the number of closest data points ($K$) based on the K-Nearest Neighbor method and the regularization parameter ($\eta$). This research optimizes the parameters $K$ and $\eta$ automatically based on the minimum error from the DRLSC classification method. The optimization method used is a hybrid of Genetic Algorithm and Particle Swarm Optimization (GAPSO). The data used for this research comes from the UCI database, specifically IRIS, WINE, and LENSA. Based on the experimental results for optimizing the DRLSC method, the fitness value produced by the GAPSO method is better than those of the Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) methods, with a fitness value difference ranging from $7.3 \times 10^{-8}$ to $0.025$.

Item Type:	Thesis (Masters)
Additional Information:	RTIf 005.1 Rir o
Uncontrolled Keywords:	Klasifikasi, Discriminatively Regularized Least Square Classification, Algoritma genetika, Particle Swarm Optimization, Pengenalan Pola, Classification, Discriminatively Regularized Least Square Classification, Genetic Algorithm, Particle Swarm Optimization, Pattern Recognition.
Subjects:	Q Science > QA Mathematics > QA402.5 Genetic algorithms. Interior-point methods.
Divisions:	Faculty of Information Technology > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	magang .
Date Deposited:	21 Sep 2026 08:14
Last Modified:	21 Sep 2026 08:14
URI:	http://repository.its.ac.id/id/eprint/144750

Actions (login required)

View Item