2017
DOI: 10.1016/j.eswa.2017.04.003
|View full text |Cite
|
Sign up to set email alerts
|

An up-to-date comparison of state-of-the-art classification algorithms

Abstract: Current benchmark reports of classification algorithms generally concern common classifiers and their variants but do not include many algorithms that have been introduced in recent years. Moreover, important properties such as the dependency on number of classes and features and CPU running time are typically not examined. In this paper, we carry out a comparative empirical study on both established classifiers and more recently proposed ones on 71 data sets originating from different domains, publicly availa… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

4
195
0
3

Year Published

2019
2019
2024
2024

Publication Types

Select...
7
1
1
1

Relationship

0
10

Authors

Journals

citations
Cited by 362 publications
(202 citation statements)
references
References 65 publications
4
195
0
3
Order By: Relevance
“…We refrain from using the fitness function (or similar optimisation criteria) to measure the manifold "quality" so as not to introduce bias towards any specific manifold learning method. The scikit-learn [14] implementation of the Random Forest (RF) classification algorithm (with 100 trees) is used as it is a widely used algorithm with high classification accuracy, is stable across a range of datasets, and has reasonably low computational cost [21]. While other algorithms could also be compared, we found the results to be generally consistent across algorithms, and so do not include these for brevity.…”
Section: Experiments Designmentioning
confidence: 99%
“…We refrain from using the fitness function (or similar optimisation criteria) to measure the manifold "quality" so as not to introduce bias towards any specific manifold learning method. The scikit-learn [14] implementation of the Random Forest (RF) classification algorithm (with 100 trees) is used as it is a widely used algorithm with high classification accuracy, is stable across a range of datasets, and has reasonably low computational cost [21]. While other algorithms could also be compared, we found the results to be generally consistent across algorithms, and so do not include these for brevity.…”
Section: Experiments Designmentioning
confidence: 99%
“…kNN is used as an example of a very simple, distance-based classification algorithm. RF, in contrast, is much more sophisticated (using an ensemble decision-based approach) and is widely used for its high classification accuracy and applicability to a wide range of datasets [44]. We use the standard default implementations of this classifiers in the scikit-learn package [34], with k = 3 for kNN, and 100 base estimators for RF.…”
Section: Classification Accuracy As a Proxy For Manifold Qualitymentioning
confidence: 99%
“…[51] can be used as a guidance. The recent study [52] collected extensive comparison of several Machine Learning algorithms. Despite not being able to contribute much to that topic, we show that for the application discussed through this paper it is indeed the case: the DNN technique by far outperforms the more classical approaches.…”
Section: B Alternative ML Techniquesmentioning
confidence: 99%