Software Open Access

Hyperparameter Optimisation for Improving Classification under Class Imbalance

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


MARC21 XML Export

<?xml version='1.0' encoding='UTF-8'?>
<record xmlns="http://www.loc.gov/MARC21/slim">
  <leader>00000nmm##2200000uu#4500</leader>
  <datafield tag="041" ind1=" " ind2=" ">
    <subfield code="a">eng</subfield>
  </datafield>
  <datafield tag="653" ind1=" " ind2=" ">
    <subfield code="a">Class Imbalance</subfield>
  </datafield>
  <datafield tag="653" ind1=" " ind2=" ">
    <subfield code="a">Hyperparameter Optimisation</subfield>
  </datafield>
  <datafield tag="653" ind1=" " ind2=" ">
    <subfield code="a">Overlapping Classes</subfield>
  </datafield>
  <controlfield tag="005">20200527202034.0</controlfield>
  <controlfield tag="001">3855193</controlfield>
  <datafield tag="711" ind1=" " ind2=" ">
    <subfield code="d">6-9 December 2019</subfield>
    <subfield code="g">SSCI</subfield>
    <subfield code="a">The 2019 IEEE Symposium Series on Computational Intelligence</subfield>
    <subfield code="c">Xiamen, China</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="u">University of Leiden</subfield>
    <subfield code="0">(orcid)0000-0002-6973-1341</subfield>
    <subfield code="a">Wojtek Kowalczyk</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="u">University of Leiden</subfield>
    <subfield code="a">Duc Anh Nguyen</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="u">Honda Research Institute Europe GmBH</subfield>
    <subfield code="a">Stefan Menzel</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="u">University of Leiden</subfield>
    <subfield code="0">(orcid)0000-0001-6768-1478</subfield>
    <subfield code="a">Thomas Bäck</subfield>
  </datafield>
  <datafield tag="856" ind1="4" ind2=" ">
    <subfield code="s">17110</subfield>
    <subfield code="z">md5:dfb7e9321f670d9127549f1012becdd8</subfield>
    <subfield code="u">https://zenodo.org/record/3855193/files/SSCI_experiment.py</subfield>
  </datafield>
  <datafield tag="542" ind1=" " ind2=" ">
    <subfield code="l">open</subfield>
  </datafield>
  <datafield tag="856" ind1="4" ind2=" ">
    <subfield code="y">Conference website</subfield>
    <subfield code="u">http://www.wikicfp.com/cfp/servlet/event.showcfp?eventid=73721</subfield>
  </datafield>
  <datafield tag="260" ind1=" " ind2=" ">
    <subfield code="c">2020-02-20</subfield>
  </datafield>
  <datafield tag="909" ind1="C" ind2="O">
    <subfield code="p">software</subfield>
    <subfield code="p">user-ecole_itn</subfield>
    <subfield code="o">oai:zenodo.org:3855193</subfield>
  </datafield>
  <datafield tag="100" ind1=" " ind2=" ">
    <subfield code="u">University of Leiden</subfield>
    <subfield code="a">Jiawen Kong</subfield>
  </datafield>
  <datafield tag="245" ind1=" " ind2=" ">
    <subfield code="a">Hyperparameter Optimisation for Improving Classification under Class Imbalance</subfield>
  </datafield>
  <datafield tag="980" ind1=" " ind2=" ">
    <subfield code="a">user-ecole_itn</subfield>
  </datafield>
  <datafield tag="536" ind1=" " ind2=" ">
    <subfield code="c">766186</subfield>
    <subfield code="a">Experience-based Computation: Learning to Optimise</subfield>
  </datafield>
  <datafield tag="540" ind1=" " ind2=" ">
    <subfield code="u">https://opensource.org/licenses/GPL-3.0</subfield>
    <subfield code="a">GNU General Public License v3.0 or later</subfield>
  </datafield>
  <datafield tag="650" ind1="1" ind2="7">
    <subfield code="a">cc-by</subfield>
    <subfield code="2">opendefinition.org</subfield>
  </datafield>
  <datafield tag="520" ind1=" " ind2=" ">
    <subfield code="a">&lt;p&gt;This is the source code used in the paper below:&lt;/p&gt;

&lt;p&gt;Jiawen Kong, Wojtek Kowalczyk, Duc Anh Nguyen, Stefan Menzel and Thomas B&amp;auml;ck, &amp;ldquo;Hyperparameter Optimisation for Improving Classification under Class Imbalance&amp;rdquo;, in 2019 IEEE Symposium Series on Computational Intelligence (SSCI), Xiamen, China, 6-9 December 2019, doi:&amp;nbsp;10.1109/SSCI44817.2019.9002679&lt;/p&gt;

&lt;p&gt;Although the class-imbalance classification problem has caught a huge amount&amp;nbsp;&lt;br&gt;
of attention, hyperparameter optimisation has not been studied in detail in&amp;nbsp;&lt;br&gt;
this field. Both classification algorithms and resampling techniques involve&amp;nbsp;&lt;br&gt;
some hyperparameters that can be tuned. This paper sets up several&amp;nbsp;&lt;br&gt;
experiments and draws the conclusion that, compared to using default&amp;nbsp;&lt;br&gt;
hyperparameters, applying hyperparameter optimisation for both&amp;nbsp;&lt;br&gt;
classification algorithms and resampling approaches can produce the best&amp;nbsp;&lt;br&gt;
results for classifying the imbalanced datasets. Moreover, this paper shows&amp;nbsp;&lt;br&gt;
that data complexity, especially the overlap between classes, has a big impact&amp;nbsp;&lt;br&gt;
on the potential improvement that can be achieved through hyperparameter&amp;nbsp;&lt;br&gt;
optimisation. Results of our experiments also indicate that using resampling&amp;nbsp;&lt;br&gt;
techniques cannot improve the performance for some complex datasets, which&amp;nbsp;&lt;br&gt;
further emphasizes the importance of analyzing data complexity before dealing&amp;nbsp;&lt;br&gt;
with imbalanced datasets.&lt;/p&gt;</subfield>
  </datafield>
  <datafield tag="773" ind1=" " ind2=" ">
    <subfield code="n">doi</subfield>
    <subfield code="i">isVersionOf</subfield>
    <subfield code="a">10.5281/zenodo.3855192</subfield>
  </datafield>
  <datafield tag="024" ind1=" " ind2=" ">
    <subfield code="a">10.5281/zenodo.3855193</subfield>
    <subfield code="2">doi</subfield>
  </datafield>
  <datafield tag="980" ind1=" " ind2=" ">
    <subfield code="a">software</subfield>
  </datafield>
</record>
18
5
views
downloads
All versions This version
Views 1818
Downloads 55
Data volume 85.5 kB85.5 kB
Unique views 1414
Unique downloads 55

Share

Cite as