Support vector machines: Their use in geotechnical engineering as illustrated using seismic liquefaction data

摘要: Abstract Empirical models based on known or measured sample data are often used to develop solutions problems in which the underlying first principles not well defined and it is possible define a concise relationship between variables, problem too complicated be described mathematically. Increasingly, various modern “learning” algorithms such as neural networks being considered that essentially map dependency inputs outputs from patterns. This study looks at fairly new pattern recognition tool support vector machines (SVM) can for solving classification-type problems. There two main ideas SVM discriminant-type The an optimum linear separating hyperplane (decision surface) separates second idea use of kernel functions (dot product vectors) apply mapping original nonlinear patterns, becomes linearly separable high-dimensional feature space. An overview presented followed by illustration its assess seismic liquefaction data. model was trained tested relatively large set comprising 226 field records performance cone penetration test measurements. overall classification success rate entire 98%.