Geri Dön

A study of a hybrid clustering using swarm intelligence techniquesand K-means algorithm

Başlık çevirisi mevcut değil.

  1. Tez No: 720741
  2. Yazar: DURDANE KOCAÇOBAN
  3. Danışmanlar: DR. XİN-SHE YANG
  4. Tez Türü: Yüksek Lisans
  5. Konular: Biyoistatistik, Biyomühendislik, Biyoteknoloji, Biostatistics, Bioengineering, Biotechnology
  6. Anahtar Kelimeler: Belirtilmemiş.
  7. Yıl: 2015
  8. Dil: İngilizce
  9. Üniversite: Middlesex University
  10. Enstitü: Yurtdışı Enstitü
  11. Ana Bilim Dalı: Belirtilmemiş.
  12. Bilim Dalı: Belirtilmemiş.
  13. Sayfa Sayısı: 77

Özet

Özet yok.

Özet (Çeviri)

The history of clustering goes back to Plato (Wikipedia, 2015). He showed a categorizing approach in his Western Philosophy by using their similar characteristics (Wikipedia, 2015). Additionally this idea was expanded and improved by Aristotle in his Categories study. On this basis, the main idea in clustering can be summarized as the categorize objects as their similarities. In data mining, the most popular clustering method is probably the K-means algorithm. The algorithm works by dividing the data points to (a predetermined number of) k clusters. While doing this partition, the purpose is to minimize intra-cluster distances and maximize inter-cluster distance. K-means is a powerful clustering method. Although there have been many clustering algorithms in the literature Kmeans clustering has always been one of popular methods. On the other hand, K-means algorithm has some flaws such as the initialization problem and trapped at local minima. K-means is very sensitive to choose centroids. Undesirable selection of K-means can affect performance of K-means. Additionally sometimes data points tend to converge at local optima related with distance measures. As a consequence, researchers have proposed many modifications or hybrids. Recently their tendency is to use swarm intelligence algorithms. In this way, there exist hybrid studies combining nature-inspired algorithms and K-means. The well-known examples of them are Particle Swarm Optimization and Firefly algorithm. In this dissertation, we will focus on both Particle Swarm Optimization and Firefly algorithm. These two algorithms have some similar characteristics and both have strong convergence. They are used for improving cluster centroids by minimizing cluster distance. In this dissertation, to improve K-means clustering algorithm, the hybrid approaches have been proposed by combining two swarm intelligence techniques (PSO and FA) with K-means. In hybrid algorithms, we evaluated to obtain the minimum error sum of squares with the help of PSO and FA, respectively. We used several benchmarking data set examples to validate the approaches and obtained three concluding remarks. Our first conclusion is that hybrid methods often obtain the much better results than classical K-means. The second conclusion is that the hybrid of PSO and K-means does not keep its performance for large data sets, while hybrid FA keeps. The third conclusion is that even though FA and K-means methods can obtain better results, but they are also the slowest method.

Benzer Tezler

  1. A hybrid swarm intelligence algorithm for simultaneous feature selection and clustering

    Eşzamanlı öznitelik seçimi ve kümeleme için hibrit bir sürü zekası algoritması

    HASAN GEREN

    Yüksek Lisans

    İngilizce

    İngilizce

    2022

    Endüstri ve Endüstri MühendisliğiOrta Doğu Teknik Üniversitesi

    Endüstri Mühendisliği Ana Bilim Dalı

    PROF. DR. NUR EVİN ÖZDEMİREL

  2. Sürü zekâsı kullanarak renkli görüntü segmentasyon tekniklerinin geliştirilmesi

    Development of color image segmentation techniques using swarm intelligence

    TAHİR SAĞ

    Doktora

    Türkçe

    Türkçe

    2015

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolSelçuk Üniversitesi

    Bilgisayar Mühendisliği Ana Bilim Dalı

    DOÇ. DR. MEHMET ÇUNKAŞ

  3. Elektrokardiyogram verilerinin iyileştirilmiş yapay arı kolonisi (MABC) algoritması ile analizi

    Analysis of electrocardiogram data by using modified artificial bee colony (MABC) algorithm

    SELİM DİLMAÇ

    Doktora

    Türkçe

    Türkçe

    2017

    Elektrik ve Elektronik Mühendisliğiİstanbul Teknik Üniversitesi

    Elektronik ve Haberleşme Mühendisliği Ana Bilim Dalı

    PROF. DR. TAMER ÖLMEZ

  4. Medical image compression based on vector quantization and discrete wavelet transform

    Vektör kuantizasyonu ve ayrık dalgacık dönüşümüne dayalı tıbbi görüntü sıkıştırma

    AZHAR ABDULHASAN MUHAMMED ALI AJAM

    Yüksek Lisans

    İngilizce

    İngilizce

    2023

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolSakarya Üniversitesi

    Bilgisayar ve Bilişim Mühendisliği Ana Bilim Dalı

    PROF. DR. AHMET ZENGİN

  5. Agrivoltaik sistemler ile elektrikli traktörleri şarj etmek için doğru arazilerin saptanması

    Determining the right lands to charge electric tractors with agrivoltaics

    SAMED PEKDEMİR

    Yüksek Lisans

    Türkçe

    Türkçe

    2023

    Enerjiİstanbul Teknik Üniversitesi

    Enerji Bilim ve Teknoloji Ana Bilim Dalı

    PROF. DR. İLHAN KOCAARSLAN