Evaluation of feature selection methods based on artificial neural network weights. Luíza da Costa, N., Dias de Lima, M., & Barbosa, R. Expert Systems with Applications, 168:114312, 2021. ![Evaluation of feature selection methods based on artificial neural network weights [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper doi abstract bibtex Weight-based feature selection (WBFS) are methods used to measure the contribution of input to output in a trained artificial neural network (ANN). Furthermore, algorithms such as Garson’s rely upon a single best neural network model or the mean importance value of several ANNs. However, different initialization weights lead to different importance values, as reported in other studies. These differences are misleading since each rank could result in different scores, altering the position of a variable in a given rank. Therefore, we propose a new methodology to assess the stability of a WBFS method. In essence, the idea is to use a voting approach to evaluate the importance of rankings. The results showed that Garson’s, Olden’s and Yoon’s algorithms are more stable methods when applied to artificial datasets. Nevertheless, its stability is considerably reduced when applied to real-world datasets. Hence, we concluded that future work should take into consideration the aforementioned instability of existing WBFS methods as applied to complex real-world data.
@article{luiza_da_costa_evaluation_2021,
title = {Evaluation of feature selection methods based on artificial neural network weights},
volume = {168},
issn = {0957-4174},
url = {https://www.sciencedirect.com/science/article/pii/S0957417420310083},
doi = {10.1016/j.eswa.2020.114312},
abstract = {Weight-based feature selection (WBFS) are methods used to measure the contribution of input to output in a trained artificial neural network (ANN). Furthermore, algorithms such as Garson’s rely upon a single best neural network model or the mean importance value of several ANNs. However, different initialization weights lead to different importance values, as reported in other studies. These differences are misleading since each rank could result in different scores, altering the position of a variable in a given rank. Therefore, we propose a new methodology to assess the stability of a WBFS method. In essence, the idea is to use a voting approach to evaluate the importance of rankings. The results showed that Garson’s, Olden’s and Yoon’s algorithms are more stable methods when applied to artificial datasets. Nevertheless, its stability is considerably reduced when applied to real-world datasets. Hence, we concluded that future work should take into consideration the aforementioned instability of existing WBFS methods as applied to complex real-world data.},
journal = {Expert Systems with Applications},
author = {Luíza da Costa, Nattane and Dias de Lima, Márcio and Barbosa, Rommel},
year = {2021},
keywords = {\#nosource},
pages = {114312},
}
Downloads: 0
{"_id":"MgEkH4EoXsgH7XFLd","bibbaseid":"luzadacosta-diasdelima-barbosa-evaluationoffeatureselectionmethodsbasedonartificialneuralnetworkweights-2021","author_short":["Luíza da Costa, N.","Dias de Lima, M.","Barbosa, R."],"bibdata":{"bibtype":"article","type":"article","title":"Evaluation of feature selection methods based on artificial neural network weights","volume":"168","issn":"0957-4174","url":"https://www.sciencedirect.com/science/article/pii/S0957417420310083","doi":"10.1016/j.eswa.2020.114312","abstract":"Weight-based feature selection (WBFS) are methods used to measure the contribution of input to output in a trained artificial neural network (ANN). Furthermore, algorithms such as Garson’s rely upon a single best neural network model or the mean importance value of several ANNs. However, different initialization weights lead to different importance values, as reported in other studies. These differences are misleading since each rank could result in different scores, altering the position of a variable in a given rank. Therefore, we propose a new methodology to assess the stability of a WBFS method. In essence, the idea is to use a voting approach to evaluate the importance of rankings. The results showed that Garson’s, Olden’s and Yoon’s algorithms are more stable methods when applied to artificial datasets. Nevertheless, its stability is considerably reduced when applied to real-world datasets. Hence, we concluded that future work should take into consideration the aforementioned instability of existing WBFS methods as applied to complex real-world data.","journal":"Expert Systems with Applications","author":[{"propositions":["Luíza","da"],"lastnames":["Costa"],"firstnames":["Nattane"],"suffixes":[]},{"propositions":["Dias","de"],"lastnames":["Lima"],"firstnames":["Márcio"],"suffixes":[]},{"propositions":[],"lastnames":["Barbosa"],"firstnames":["Rommel"],"suffixes":[]}],"year":"2021","keywords":"#nosource","pages":"114312","bibtex":"@article{luiza_da_costa_evaluation_2021,\n\ttitle = {Evaluation of feature selection methods based on artificial neural network weights},\n\tvolume = {168},\n\tissn = {0957-4174},\n\turl = {https://www.sciencedirect.com/science/article/pii/S0957417420310083},\n\tdoi = {10.1016/j.eswa.2020.114312},\n\tabstract = {Weight-based feature selection (WBFS) are methods used to measure the contribution of input to output in a trained artificial neural network (ANN). Furthermore, algorithms such as Garson’s rely upon a single best neural network model or the mean importance value of several ANNs. However, different initialization weights lead to different importance values, as reported in other studies. These differences are misleading since each rank could result in different scores, altering the position of a variable in a given rank. Therefore, we propose a new methodology to assess the stability of a WBFS method. In essence, the idea is to use a voting approach to evaluate the importance of rankings. The results showed that Garson’s, Olden’s and Yoon’s algorithms are more stable methods when applied to artificial datasets. Nevertheless, its stability is considerably reduced when applied to real-world datasets. Hence, we concluded that future work should take into consideration the aforementioned instability of existing WBFS methods as applied to complex real-world data.},\n\tjournal = {Expert Systems with Applications},\n\tauthor = {Luíza da Costa, Nattane and Dias de Lima, Márcio and Barbosa, Rommel},\n\tyear = {2021},\n\tkeywords = {\\#nosource},\n\tpages = {114312},\n}\n\n\n\n","author_short":["Luíza da Costa, N.","Dias de Lima, M.","Barbosa, R."],"key":"luiza_da_costa_evaluation_2021","id":"luiza_da_costa_evaluation_2021","bibbaseid":"luzadacosta-diasdelima-barbosa-evaluationoffeatureselectionmethodsbasedonartificialneuralnetworkweights-2021","role":"author","urls":{"Paper":"https://www.sciencedirect.com/science/article/pii/S0957417420310083"},"keyword":["#nosource"],"metadata":{"authorlinks":{}},"html":""},"bibtype":"article","biburl":"https://bibbase.org/zotero/fsimonetta","dataSources":["pzyFFGWvxG2bs63zP"],"keywords":["#nosource"],"search_terms":["evaluation","feature","selection","methods","based","artificial","neural","network","weights","luíza da costa","dias de lima","barbosa"],"title":"Evaluation of feature selection methods based on artificial neural network weights","year":2021}