Experimental Study: Comparison of clustering algorithms

Abstract

One of the most important processes in the machine learning is the clustering. The clustering is an unsupervised process that gathers all similar measurements to identify and put them in groups based on specific measurements. Clustering task is required in many applications such as, text analysis, data visualization, nature language processing, image processing, computer vision, and even gene expression analysis. This work tends to make a comparison study to analyze the performance of different clustering algorithms using different datasets. We conduct some experimental results to evaluate the effectiveness of six clustering algorithms: hard K mean, fuzzy K mean, Locality weighted of hard K mean, Locality weighted of fuzzy K mean, Hierarchical , and DBSCAN algorithms. We use synaptic and real dataset in our experiments. We synthesize three different datasets to analyze the performance: imbalanced classes dataset, an outlier dataset, and moon dataset. Additionally, we perform image segmentation and compression using these clustering algorithms. Finally, we test the performance of the algorithms by performing facial expression clustering, which is one of the most challenging problem in the computer vision.

Authors and Affiliations

Mohammed Dawod, Mays Hasan, Amar Daood

Keywords

Related Articles

Green Buildings: The Future

Due to increase in green house gas emission and increasing climatic changes, the building architecture needs to change to keep up with the time. Green buildings are the future buildings. They utilize natural resources fo...

Simultaneous Computing of Sets of Maximally Different Alternatives to Optimal Solutions

In solving many “real world” engineering optimization applications, it is generally preferable to formulate several quantifiably good alternatives that provide distinct perspectives to the particular problem. This is bec...

Soft Computing Technique for the Control of Triple-Lift Luo Converter

Positive output Luo converters are a series of new DC-DC step-up (boost) converters, which were developed from prototypes using voltage lift technique. These converters perform positive to positive DC-DC voltage increasi...

Experimental studies on fiber reinforced concrete

The concepts of using fibres in order to reinforce matrices weak in tension is more than 4500 years old.since Portland cement concrete started to be used widely as a construction material attempts were made to use fibres...

Comparison of Mechanical Properties of Bare and Stellite-21 Coated Aisi 316l, Sprayed By D-Gun Process

Present paper describes the mechanical properties of Stellite-21 coated AISI- 316LStainless steel. Specimens were coated using Detonation Gun thermal spray process,with different coating thicknesses of Stellite-21 rangin...

Download PDF file
  • EP ID EP392042
  • DOI 10.9790/9622-0708042334.
  • Views 109
  • Downloads 0

How To Cite

Mohammed Dawod, Mays Hasan, Amar Daood (2017). Experimental Study: Comparison of clustering algorithms. International Journal of engineering Research and Applications, 7(8), 23-34. https://europub.co.uk/articles/-A-392042