Analytic study on the performance of multi-classification approaches in case-based reasoning systems: Medical data exploration. Bastidas, D., Piñeros, C., Peluffo-Ordóñez, D., H., Sierra, L., M., Becerra, M., A., & Umaquinga-Criollo, A., C. RISTI - Revista Iberica de Sistemas e Tecnologias de Informacao, 2020.
Analytic study on the performance of multi-classification approaches in case-based reasoning systems: Medical data exploration [link]Website  abstract   bibtex   5 downloads  
This paper compares the main combinations of classifiers (Sequential, Parallel and Stacking) over two remarkable medical data collections: Cleveland and Dermatology. The principal rationale underlying the use of multiple classifiers is that together the methods may be powered rather than their individual behavior. Such a premise is validated through the identification of the best the combination reaching the lowest error rate within a case-based reasoning system (CBR). The different combinations are essentially formed by five different classifiers greatly different regarding their nature and inception: SVM (Support Vector Machines), Parzen, Random Forest, K-NN (k-nearest neighbors) and Naive Bayes. From experimental results, it can be inferred that the combination of techniques is greatly useful. Also, in this work, some key aspects and hints are discussed about the relationship between the nature of the input data and the classification (either individual or mixture of classifiers) stage building within a CBR framework.
@article{
 title = {Analytic study on the performance of multi-classification approaches in case-based reasoning systems: Medical data exploration},
 type = {article},
 year = {2020},
 keywords = {Case based reasoning,Classifiers fusion,Dermatology disease,Heart disease},
 websites = {https://search.proquest.com/docview/2451420129/fulltextPDF/F4AF5E590BD14D5EPQ/9},
 id = {110b6f27-3135-3732-a220-25368f4c99bb},
 created = {2021-05-10T23:31:46.313Z},
 file_attached = {false},
 profile_id = {aba9653c-d139-3f95-aad8-969c487ed2f3},
 group_id = {b8c873a2-ed58-396e-b124-8a273665874f},
 last_modified = {2021-05-10T23:31:46.313Z},
 read = {false},
 starred = {false},
 authored = {false},
 confirmed = {true},
 hidden = {false},
 citation_key = {Bastidas2020},
 private_publication = {false},
 abstract = {This paper compares the main combinations of classifiers (Sequential, Parallel and Stacking) over two remarkable medical data collections: Cleveland and Dermatology. The principal rationale underlying the use of multiple classifiers is that together the methods may be powered rather than their individual behavior. Such a premise is validated through the identification of the best the combination reaching the lowest error rate within a case-based reasoning system (CBR). The different combinations are essentially formed by five different classifiers greatly different regarding their nature and inception: SVM (Support Vector Machines), Parzen, Random Forest, K-NN (k-nearest neighbors) and Naive Bayes. From experimental results, it can be inferred that the combination of techniques is greatly useful. Also, in this work, some key aspects and hints are discussed about the relationship between the nature of the input data and the classification (either individual or mixture of classifiers) stage building within a CBR framework.},
 bibtype = {article},
 author = {Bastidas, David and Piñeros, Camilo and Peluffo-Ordóñez, Diego H. and Sierra, Luz Marina and Becerra, Miguel A. and Umaquinga-Criollo, Ana C.},
 journal = {RISTI - Revista Iberica de Sistemas e Tecnologias de Informacao}
}

Downloads: 5