Repositório Comunidade:http://comum.rcaap.pt/handle/123456789/26192015-07-31T17:32:48Z2015-07-31T17:32:48ZFeatures selection in Discrete Discriminant AnalysisMarques, AnabelaFerreira, Ana SousaCardoso, Margaridahttp://comum.rcaap.pt/handle/123456789/81342015-03-25T21:26:52Z2011-06-01T00:00:00ZTítulo: Features selection in Discrete Discriminant Analysis
Autor: Marques, Anabela; Ferreira, Ana Sousa; Cardoso, Margarida
Resumo: In discrete discriminant analysis dimensionality problems occur, particularly when dealing with data from the social sciences, humanities and health.
In these domains, one often has to classify entities with a high number of explanatory variables when compared to the number of observations available.
In the present work we address the problem of features selection in classification, aiming to identify the variables that most discriminate between the a priori defined classes, reducing the number of parameters to estimate, turning the results easier to interpret and reducing the runtime of the methods used. We specially address classification using a recently methodological approach based on a linear combination of the First-order Independence Model (FOIM) and the Dependence Trees Model (DTM).
Data of small and moderate size are considered.
Descrição: Resumo de comunicação em póster apresentada em 14th International Conference on Applied Stochastic Models and Data Analysis (ASMDA2011), Rome, June 7-10 20112011-06-01T00:00:00ZCombining models in discrete discriminant analysisMarques, AnabelaFerreira, Ana SousaCardoso, Margaridahttp://comum.rcaap.pt/handle/123456789/81312015-03-25T18:31:57Z2010-01-01T00:00:00ZTítulo: Combining models in discrete discriminant analysis
Autor: Marques, Anabela; Ferreira, Ana Sousa; Cardoso, Margarida
Resumo: Diverse Discrete Discriminant Analysis (DDA) models perform differently on different sample observations (Brito et al. (2006)). This fact has encouraged research in combined models for DDA. This research seems to be specially promising when the a priori
classes are not well separated or when small or moderate sized samples are considered, which often occurs in practice. In this work we evaluate the performance of a linear combination of two DDA models (Marques et
al. (2008)): the First-Order Independence Model (FOIM) and the Dependence Trees Model (DTM) (Celeux and Nakache (1994). The pro-
posed methodology also uses a Hierarchical Coupling Model (HIERM) when addressing multiclass classification problems, decomposing the multiclass problems into several bi-class problems, using a binary tree structure (Sousa Ferreira (2000)). The analysis is based both on simulated and real datasets. Results include measures of precision regarding a training set, a test set and cross-validation. The R software is used for the algorithm's implementation.
Descrição: Resumo da comunicação em póster apresentada em International Conference on Trends and Perspectives in Linear Statistical Inference (LinStat'2010), Tomar, Portugal, 27-31 July, 20102010-01-01T00:00:00ZDiscrete discriminant analysis: The performance of combining modelsMarques, AnabelaFerreira, Ana SousaCardoso, Margaridahttp://comum.rcaap.pt/handle/123456789/81292015-03-25T17:13:28Z2009-03-01T00:00:00ZTítulo: Discrete discriminant analysis: The performance of combining models
Autor: Marques, Anabela; Ferreira, Ana Sousa; Cardoso, Margarida
Resumo: The idea of combinig models in Discrete Discriminant Analysis (DDA) is present in a growing number of papers which aim to obtain more robust and more stable models than any of the competing ones. This seems to be a promising approach since it is known that different DDA models perform differently on different subjects [1]. This is a particularly relevant issue when the groups are not well separeted, which often occurs in practice. recently, a new methodological approach was proposed based on a linear combination of the First-order Independence Model (FOIM) and the Dependence trees Model (DTM) ([3] and [2]). In the present work we further explore the referred approach. Since FOIM assumes that the P discrete predictive variables are independent in each group and DTM takes the predictors relationships into account, we think that the proposed approach could be sucessfully applied to many real situations. In order to evaluate its performance, we consider both real and simulated data. Furthermore we present comparisons with alternative models performance. According to the training sample size the leave-one-out approach, v-fold cross validation or assessing the error rate in a test sample are considered. The MATLAB software is used for the algorithms' implementation.
Descrição: Resumo de comunicação oral apresentada em 11th Conference of the International Federation of Classification Societes (IFCS 2009), Dresden, Germany, 13-18 March 20092009-03-01T00:00:00ZCombining models in discrete discriminant analysis in the multiclass caseMarques, AnabelaFerreira, Ana SousaCardoso, Margaridahttp://comum.rcaap.pt/handle/123456789/81272015-03-25T16:10:10Z2008-01-01T00:00:00ZTítulo: Combining models in discrete discriminant analysis in the multiclass case
Autor: Marques, Anabela; Ferreira, Ana Sousa; Cardoso, Margarida
Resumo: The idea of combining models in Discrete Discriminant Analysis (DDA) is present in a growing number of papers which aim to obtain more robust and more stable models than any of the competing ones. This seems to be a promising approach since it is known that different DDA models perform differently on different subjects (Brito et al.(2006)). In particular, this will be a more relevant issue if the groups are not well separated, which often occurs in practice.
In the present work a new methodological approach is suggested which is based on DDA models' combination. The multiclass problem is decomposed into several dichotomous problems that are nested in a hierarchical binary tree (Sousa Ferreira (2000), Brito et al. (2006)) and at each level of the binary tree a new combining model is proposed to derive the decision rule. This combining model is based on two well known models in the literature - the First-order Independence Model (FOIM) and the Dependence Trees Model (DTM) (Celeux and Nakache (1994)).
The MATLAB software is used for the algorithms' implementation and the proposed
approach is illustrated in a DDA application.
Descrição: Resumo de comunicação oral em póster apresentado em COMPSTAT2008 - 18th International Conference on Computational Statistics, Porto, Portugal, 24 a 29 de Agosto 20082008-01-01T00:00:00Z