Deformation Prediction Method Based On Support Vector Machines

Since the twenty-first century, as science and technology and the continuous development of industrial mass production, but also with the rapid population growth, which requires material and energy production must be accelerated to meet the current needs of social development. At the same time, leding to the increasing large scale of the various projects, but also put forward a higher demand in the size and shape of modern building (structures),while the cost is more and more high, the requirements are more and more precise.As we know, construction (structures) in the construction and operation period, due to a variety of subjective and objective factors will have some deformations, such deformation, if exceeding the limits of the relevant orders, that will affect the normal use of the building, and even would endanger the safety of buildings and bring great losses to the society and people's lives. Therefore, if we can timely learn the deformation information of the construction (structures), and change the projection maneuvers or take the effective emergency mothed according to its dynamic changes, we can reduce the disastrous due to the deformation of the building.Currently, it has been increasingly focused on the job of using the state monitoring data to analytize the deformation of the construction (structures). Especially in enhancing the monitoring and forecasting analysis in important projects, such as large-scale powerhouses, dams, urban high-rise buildings and some important military buildings and facilities..With the rapid development of computer hardware and software technology, many scholars began to introduce the machine learning into the building deformation prediction, and establish many new models, such as the Artificial Neural Network (ANN) which has been widely applied to complex systems modeling.In this paper, through using support vector machine to build regression models, and on this basis using prectical monitoring data to predict deformation of the building. Support vector machine (Support Vector Machine, SVM) as a new method of machine learning,which is put forward by Vapnik and others in 1995,as a new method on pattern recognition which is based on the statistical learning theory, in recent years,the theoretical study and using implementation have been made a big breakthrough and began to an effective means to overcome the "curse of dimension" and "over learning". SVM has a simple mathematical form, and has all optimality, and good generalization ability, which has been a valid tool in using in classification, regression, probability density estimation,and in recent years, it has been made successful using in object recognition, text classification, function approximation and time series prediction and many other practical applications. and in the paper, the author takes advantage of the prediction results of SVM to compare with the prediction results BP neural networks, in order to verify support vector machine model which can be good in using in prediction of deformation, and support vector machine prediction in deformation also has a good advantage.