Automated selection of interaction effects in sparse kernel methods to predict pregnancy viability

作者: Vanya Van Belle , Paulo Lisboa

DOI: 10.1109/CIDM.2013.6597213

关键词: Predictive modellingRadial basis function kernelKernel methodMathematical modelArtificial intelligencePolynomial kernelSupport vector machineSelection (genetic algorithm)Computer scienceMachine learningData miningInterpretation (logic)

摘要: Support vector machines are highly flexible and generalizing mathematical models that can be used to build prediction models. Their success on a field is not followed by their application in practice due black-box nature. The RBF kernel often but the good performance cannot accompanied an interpretation of results. We present method visualize different components propose select relevant ones. proposed able automatically detect important main two-way interaction effects while still obtaining interpretable illustrated large dataset predict viability pregnancies at end first trimester based initial scan findings.

参考文章(16)
Marc G. Genton, Classes of kernels for machine learning: a statistics perspective international conference on artificial intelligence and statistics. ,vol. 2, pp. 299- 312 ,(2002) , 10.5555/944790.944815
J.A.K. Suykens, J. Vandewalle, Least Squares Support Vector Machine Classifiers Neural Processing Letters. ,vol. 9, pp. 293- 300 ,(1999) , 10.1023/A:1018628609742
O. L. Mangasarian, Paul S. Bradley, Feature Selection via Concave Minimization and Support Vector Machines international conference on machine learning. pp. 82- 90 ,(1998)
K. Pelckmans, B. De Moor, J. A. K. Suykens, Componentwise support vector machines for structure detection international conference on artificial neural networks. ,vol. 3697, pp. 643- 648 ,(2005) , 10.1007/11550907_102
Juned Siddique, Ofer Harel, MIDAS: A SAS macro for multiple imputation using distance-aided selection of donors Journal of Statistical Software. ,vol. 29, pp. 1- 18 ,(2009) , 10.18637/JSS.V029.I09
S. Xavier-de-Souza, J.A.K. Suykens, J. Vandewalle, D. Bolle, Coupled Simulated Annealing systems man and cybernetics. ,vol. 40, pp. 320- 335 ,(2010) , 10.1109/TSMCB.2009.2020435
Vanya M. C. A. Van Belle, Ben Van Calster, Dirk Timmerman, Tom Bourne, Cecilia Bottomley, Lil Valentin, Patrick Neven, Sabine Van Huffel, Johan A. K. Suykens, Stephen Boyd, A Mathematical Model for Interpretable Clinical Decision Support with Applications in Gynecology PLoS ONE. ,vol. 7, pp. e34312- ,(2012) , 10.1371/JOURNAL.PONE.0034312
Cecilia Bottomley, Vanya Van Belle, Emma Kirk, Sabine Van Huffel, Dirk Timmerman, Tom Bourne, Accurate prediction of pregnancy viability by means of a simple scoring system Human Reproduction. ,vol. 28, pp. 68- 76 ,(2013) , 10.1093/HUMREP/DES352
Leo Breiman, Better subset regression using the nonnegative garrote Technometrics. ,vol. 37, pp. 373- 384 ,(1995) , 10.2307/1269730
Emmanuel J. Candès, Michael B. Wakin, Stephen P. Boyd, Enhancing Sparsity by Reweighted ℓ 1 Minimization Journal of Fourier Analysis and Applications. ,vol. 14, pp. 877- 905 ,(2008) , 10.1007/S00041-008-9045-X