Centralized vs. distributed feature selection methods based on data complexity measures

Morán-Fernández, Laura; Bolón-Canedo, Verónica; Alonso‐Betanzos, Amparo

doi:10.1016/j.knosys.2016.09.022

Cited by 62 publications

(32 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Following the recommendations in [32], the vertical approach has the drawback to not handle redundant features. Indeed, with the vertical partition, the features were distributed across the packets thereby; it will be more difficult to detect redundancy between them.…”

Section: Analysis and Discussionmentioning

confidence: 99%

“…The drawback of this methodology was its dependence on the classifier used. More recently, the same authors propose a distributed approach based on data complexity measures [32], this method was carried out for both the horizontal and the vertical technique. To combine the partial outputs obtained from feature selection algorithm applied to each subset, a merging process using the theoretical complexity is applied to these feature subsets.…”

Section: Related Workmentioning

confidence: 99%

“…Following the recommendations in [32], the use of the distributed approaches is recommended instead of the traditional centralized methods when dealing with large datasets. On the other hand, the vertical distribution is recommended in case that better classification performance is more important than a smaller storage requirement.…”

Section: Related Workmentioning

confidence: 99%

“…For testing the adequacy of our proposal, we select four widely used classifiers: C4.5, naïve Bayes, k-Nearest Neighbor and Support Vector Machine, more details are described in the previous section. We also compare our proposal with the horizontal and the vertical strategies proposed in [32], as well as the centralized approach in terms of number of selected features and classification accuracy. Furthermore, for the distributed approaches, it is mandatory to choose the adequate threshold to combine the partial outputs.…”

Section: Data and Toolsmentioning

confidence: 99%

“…As can be seen, the number of features selected by the vertical approaches was larger than those selected by our distributed methods as well as all others approaches. The reason behind is that, with the vertical partition the features were distributed across the packets, so detecting redundancy between features will be more difficult if they were in different partitions [32]. Whereas, there is no significant differences between the number of features selected by vertical and the centralized approach.…”

Section: Threshold Impactmentioning

confidence: 99%

See 4 more Smart Citations

A New Horizo-Vertical Distributed Feature Selection Approach

Zerhari

Lehcen

Mouline

2018

Cybernetics and Information Technologies

View full text Add to dashboard Cite

Feature selection technique has been a very active research topic that addresses the problem of reducing the dimensionality. Whereas, datasets are continuously growing over time both in samples and features number. As a result, handling both irrelevant and redundant features has become a real challenge. In this paper we propose a new straightforward framework which combines the horizontal and vertical distributed feature selection technique, called Horizo-Vertical Distributed Feature Selection approach (HVDFS), aimed at achieving good performances as well as reducing the number of features. The effectiveness of our approach is demonstrated on three well-known datasets compared to the centralized and the previous distributed approach, using four well-known classifiers.

show abstract

Section: Analysis and Discussionmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Data and Toolsmentioning

confidence: 99%

Section: Threshold Impactmentioning

confidence: 99%

See 3 more Smart Citations

A New Horizo-Vertical Distributed Feature Selection Approach

Zerhari

Lehcen

Mouline

2018

Cybernetics and Information Technologies

View full text Add to dashboard Cite

show abstract

HMDE‐FS: A homogeneous distributed ensemble feature selection framework based on resampling with/without replacement

Nosrati

Rahmani

2023

Concurrency and Computation

View full text Add to dashboard Cite

Summary Applying ensemble feature selection (EFS) models in various problems has not been actively discussed, and there has been a lack of effort to make it applicable in the situations such as distributed environments. Due to restrictions of centralized algorithms such as their poor scalability in the high dimension data and also distributed nature of some data, using the traditional centralized computing for dealing with such problems may be inevitable. This paper aims to develop a homogenous distributed ensemble feature selection (HMDE‐FS) framework through a distributed resampling approach rather than a centralized one. The homogenous ensembles mainly operate along with a resampling process, so applying various methods to resampling can affect the performance of the model. Among various strategies, those with and without replacement are two of the main technique families, hence we investigated the efficiency of two well‐known with/without replacement techniques: bootstrapping (BS) and cross‐validation (CV) inspired method that we named crisscross (CC). The proposed HMDE‐FS approaches are tested on eight datasets, and the heavy experimental results illustrate that these methods considerably reduce runtime, while classification accuracy maintains its competitiveness.

show abstract

Ensemble feature selection for multi‐label text classification: An intelligent order statistics approach

Miri

Dowlatshahi

Hashemi

et al. 2022

Int J of Intelligent Sys

View full text Add to dashboard Cite

Because of the overgrowth of data, especially in text format, the value and importance of multi‐label text classification have increased. Aside from this, preprocessing and particularly intelligent feature selection (FS) are the most important step in classification. Each FS finds the best features based on its approach, but we try to use a multi‐strategy approach to find more useful features. Evaluating and comparing features’ importance and relevance makes using multiple strategy and methods more suitable than conventional approaches because each feature is measured based on several perspectives. Nevertheless, the ensemble FS merges the final performance results of various methods to take advantage of different methods’ strengths and better classify. In this article, we have proposed an ensemble FS method for multi‐label text data (MLTD) for the first time using the order statistics (EMFS) approach. We have utilized four multi‐label FS (MLFS) algorithms with various particular performances to achieve a good result. In this method, as one of the most important statistics methods, Order Statistics was used to aggregate the ranks of different algorithms, which is robust against noise, redundant and inessential features. In the end, the performance of EMFS, executing six MLTDs, was evaluated according to six performance criteria (ranking‐based and classification‐based). Surprisingly, the proposed method was more accurate than others among all used MLTDs. The proposed method has improved by 1.5% compared to other methods. This value is based on the results obtained based on six evaluation criteria and all tested data sets.

show abstract

Centralized vs. distributed feature selection methods based on data complexity measures

Cited by 62 publications

References 11 publications

A New Horizo-Vertical Distributed Feature Selection Approach

A New Horizo-Vertical Distributed Feature Selection Approach

HMDE‐FS: A homogeneous distributed ensemble feature selection framework based on resampling with/without replacement

Ensemble feature selection for multi‐label text classification: An intelligent order statistics approach

Contact Info

Product

Resources

About