Software Open Access

Hyperparameter Optimisation for Improving Classification under Class Imbalance

Jiawen Kong; Wojtek Kowalczyk; Duc Anh Nguyen; Stefan Menzel; Thomas Bäck


JSON-LD (schema.org) Export

{
  "inLanguage": {
    "alternateName": "eng", 
    "@type": "Language", 
    "name": "English"
  }, 
  "description": "<p>This is the source code used in the paper below:</p>\n\n<p>Jiawen Kong, Wojtek Kowalczyk, Duc Anh Nguyen, Stefan Menzel and Thomas B&auml;ck, &ldquo;Hyperparameter Optimisation for Improving Classification under Class Imbalance&rdquo;, in 2019 IEEE Symposium Series on Computational Intelligence (SSCI), Xiamen, China, 6-9 December 2019, doi:&nbsp;10.1109/SSCI44817.2019.9002679</p>\n\n<p>Although the class-imbalance classification problem has caught a huge amount&nbsp;<br>\nof attention, hyperparameter optimisation has not been studied in detail in&nbsp;<br>\nthis field. Both classification algorithms and resampling techniques involve&nbsp;<br>\nsome hyperparameters that can be tuned. This paper sets up several&nbsp;<br>\nexperiments and draws the conclusion that, compared to using default&nbsp;<br>\nhyperparameters, applying hyperparameter optimisation for both&nbsp;<br>\nclassification algorithms and resampling approaches can produce the best&nbsp;<br>\nresults for classifying the imbalanced datasets. Moreover, this paper shows&nbsp;<br>\nthat data complexity, especially the overlap between classes, has a big impact&nbsp;<br>\non the potential improvement that can be achieved through hyperparameter&nbsp;<br>\noptimisation. Results of our experiments also indicate that using resampling&nbsp;<br>\ntechniques cannot improve the performance for some complex datasets, which&nbsp;<br>\nfurther emphasizes the importance of analyzing data complexity before dealing&nbsp;<br>\nwith imbalanced datasets.</p>", 
  "license": "https://opensource.org/licenses/GPL-3.0", 
  "creator": [
    {
      "affiliation": "University of Leiden", 
      "@type": "Person", 
      "name": "Jiawen Kong"
    }, 
    {
      "affiliation": "University of Leiden", 
      "@id": "https://orcid.org/0000-0002-6973-1341", 
      "@type": "Person", 
      "name": "Wojtek Kowalczyk"
    }, 
    {
      "affiliation": "University of Leiden", 
      "@type": "Person", 
      "name": "Duc Anh Nguyen"
    }, 
    {
      "affiliation": "Honda Research Institute Europe GmBH", 
      "@type": "Person", 
      "name": "Stefan Menzel"
    }, 
    {
      "affiliation": "University of Leiden", 
      "@id": "https://orcid.org/0000-0001-6768-1478", 
      "@type": "Person", 
      "name": "Thomas B\u00e4ck"
    }
  ], 
  "url": "https://zenodo.org/record/3855193", 
  "datePublished": "2020-02-20", 
  "@type": "SoftwareSourceCode", 
  "keywords": [
    "Class Imbalance", 
    "Hyperparameter Optimisation", 
    "Overlapping Classes"
  ], 
  "@context": "https://schema.org/", 
  "identifier": "https://doi.org/10.5281/zenodo.3855193", 
  "@id": "https://doi.org/10.5281/zenodo.3855193", 
  "workFeatured": {
    "url": "http://www.wikicfp.com/cfp/servlet/event.showcfp?eventid=73721", 
    "alternateName": "SSCI", 
    "location": "Xiamen, China", 
    "@type": "Event", 
    "name": "The 2019 IEEE Symposium Series on Computational Intelligence"
  }, 
  "name": "Hyperparameter Optimisation for Improving Classification under Class Imbalance"
}
19
5
views
downloads
All versions This version
Views 1919
Downloads 55
Data volume 85.5 kB85.5 kB
Unique views 1515
Unique downloads 55

Share

Cite as