FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W212669514> ?p ?o ?g. }

• W212669514 abstract "Many bioinformatics applications, such as the microarray gene expression data modeling, are characterized by a large number of variables (e.g. genes) and small number of samples. Finding the most suitable set of variables for the modeling task (e.g. classification, prediction, profiling) is an important procedure, as it will define to a greater extent the accuracy of the model and its applicability. Several methods have been used so far, that include: correlation analysis; signal to noise ratio analysis; t-test [1-6]. The approach proposed here is based on a supervised clustering, achieved in an evolving connectionist system (ECOS)[7]. The ECOS is developed as a classification model on gene expression data. It learns to classify the data into various classes. The classification model is based on clustering and the formed clusters are used to select the sets of most important variables for each of them. The proposed method applies ECOS on the whole variable space (e.g. 7,129 genes) and using the formed clusters selects the most significant genes for the task. As a case study, we have used the lymphoma microarray data [5] which consists of 77 samples classified into two classes Diffuse Large B-cell lymphoma (DLBCL) and Follicular lymphoma (FL). ECOS are evolving connectionist systems that adapt their structure and functionality according to the data. An ECOS for classification clusters in a supervised way vector samples according to their ‘similarity’ in the Euclidean problem space through adjusting an influence field for each cluster. For the case study here, the whole data set of microarray lymphoma data is fed into the system, distances between the samples are calculated and rule nodes (prototypes, cluster centers) are created to cluster (partition) the input space of 7,129 genes by their expression values. The feature selection procedure consists of applying the signal to noise ratio method, not on the pair of data from the two classes, but on all possible combinations of pairs of clusters from each class. In our experiment there were 22 clusters for the first class and 10 for the second class of lymphoma data formed on the whole 7,129 gene expression data space. After the selection of a set of variables for each cluster and the union of these sets, a new ECOS was then trained on the selected variable set and tested through the leave one out method. The result is 89.61% accuracy. These results outperformed an ECOS trained on a selected set of variables with the use of the signal to noise ratio method applied in the traditional way – accuracy of 84.95%. The proposed method combines the task of classification and the task of feature selection and uses the obtained clusters in an ECOS that learns a classification task to extract specific features for each of the clusters of class data. The method overcomes the problem of the signal to noise ratio method when data of the same class are spread in several clusters of the problem space and overlap with clusters of another classes." @default.
• W212669514 created "2016-06-24" @default.
• W212669514 creator A5009354879 @default.
• W212669514 creator A5024893012 @default.
• W212669514 date "2002-01-01" @default.
• W212669514 modified "2023-09-27" @default.
• W212669514 title "A Novel Feature Selection Method using Evolving Supervised Clustering and Applications for Gene Expression Data Modeling" @default.
• W212669514 cites W1487018788 @default.
• W212669514 cites W2064208261 @default.
• W212669514 cites W2088085204 @default.
• W212669514 cites W2088851040 @default.
• W212669514 cites W2128985829 @default.
• W212669514 cites W2137476312 @default.
• W212669514 cites W2147246240 @default.
• W212669514 hasPublicationYear "2002" @default.
• W212669514 type Work @default.
• W212669514 sameAs 212669514 @default.
• W212669514 citedByCount "0" @default.
• W212669514 crossrefType "journal-article" @default.
• W212669514 hasAuthorship W212669514A5009354879 @default.
• W212669514 hasAuthorship W212669514A5024893012 @default.
• W212669514 hasConcept C119857082 @default.
• W212669514 hasConcept C120174047 @default.
• W212669514 hasConcept C124101348 @default.
• W212669514 hasConcept C148483581 @default.
• W212669514 hasConcept C153180895 @default.
• W212669514 hasConcept C154945302 @default.
• W212669514 hasConcept C41008148 @default.
• W212669514 hasConcept C73555534 @default.
• W212669514 hasConceptScore W212669514C119857082 @default.
• W212669514 hasConceptScore W212669514C120174047 @default.
• W212669514 hasConceptScore W212669514C124101348 @default.
• W212669514 hasConceptScore W212669514C148483581 @default.
• W212669514 hasConceptScore W212669514C153180895 @default.
• W212669514 hasConceptScore W212669514C154945302 @default.
• W212669514 hasConceptScore W212669514C41008148 @default.
• W212669514 hasConceptScore W212669514C73555534 @default.
• W212669514 hasLocation W2126695141 @default.
• W212669514 hasOpenAccess W212669514 @default.
• W212669514 hasPrimaryLocation W2126695141 @default.
• W212669514 hasRelatedWork W1228375242 @default.
• W212669514 hasRelatedWork W1584722874 @default.
• W212669514 hasRelatedWork W1859349462 @default.
• W212669514 hasRelatedWork W2057518933 @default.
• W212669514 hasRelatedWork W2080736917 @default.
• W212669514 hasRelatedWork W2112059892 @default.
• W212669514 hasRelatedWork W2115290533 @default.
• W212669514 hasRelatedWork W2118076212 @default.
• W212669514 hasRelatedWork W2123283518 @default.
• W212669514 hasRelatedWork W2127912512 @default.
• W212669514 hasRelatedWork W2314597787 @default.
• W212669514 hasRelatedWork W2335756805 @default.
• W212669514 hasRelatedWork W2533593624 @default.
• W212669514 hasRelatedWork W2535976114 @default.
• W212669514 hasRelatedWork W2772740008 @default.
• W212669514 hasRelatedWork W2950067894 @default.
• W212669514 hasRelatedWork W3023267841 @default.
• W212669514 hasRelatedWork W3155563246 @default.
• W212669514 hasRelatedWork W2183576892 @default.
• W212669514 hasRelatedWork W2869977022 @default.
• W212669514 isParatext "false" @default.
• W212669514 isRetracted "false" @default.
• W212669514 magId "212669514" @default.
• W212669514 workType "article" @default.