Reasearch On Sparse Least Squares Support Vector Machines For Large-scale Data Samples

The least squares support vector machine algorithm is improved from the classical support vector machine algorithm,and the error squared term is introduced,which makes the inequality constraint problem in the classical support vector machine algorithm become an equality constraint problem,but at the same time because of the improvement,it is solved.During the process,all data sample points in the least squares srupport vector machine model are used as a model of the Lagrange multiplier to play a role in the establishment of the model,which makes the least squares support vector machine cannot be used in large-scale data concentrated.However,in practical applications,the data sample set often contains a large amount of data,and the data of different data sample sets have different characteristics.Therefore,how to make the least squares support vector machine applicable to complex and diverse large-scale data samples The main research content of this topic is in the paper.The main work of this paper is as follows:1)Through the analysis of the least squares support vector machine algorithm process,not all data sample points play a key role in the establishment of the model,but it is different from the data close to the classification decision surface in the support vector machine algorithm.The sample point has a large contribution to the establishment of the model,and the least squares support vector machine algorithm plays a key role in the establishment of the model.It is the most recent and very far data sample point from the classification decision surface.2)Introducing a clustering algorithm to select data sample points for the establishment of a least squares support vector machine model in complex and diverse large-scale data samples.The characteristics of the K-means clustering algorithm and the mean shift clustering algorithm The comparative study of the characteristics of the algorithm,the final selection of the mean shift clustering algorithm to select the data sample points that play a key role in the establishment of the least squares support vector machine model,using a new subset of data samples for model training,and in the real data Set tests to achieve optimization and improvement of the model.3)In order to verify the effectiveness of the proposed algorithm,the least squares support vector machine model reduced by the mean shift clustering algorithm is applied to the actual industrial projects.For the problem that the temperature of the seeding stage cannot be controlled automatically during the growth process of Czochralski silicon single crystal,this paper proposes a method of automatic temperature detection and adjustment based on the aperture image.The aperture original image is acquired,image processing is performed,and the aperture image is classified using a mean-shifting clustering algorithm and a reduced least square support vector machine.The model output is the temperature mode to which the aperture image belongs,thereby realizing the automatic detection of the crystal temperature.And identification.The experiment of crystal growth shows that the classification method of the reduced least squares support vector machine model proposed by the mean shift clustering algorithm in this paper realizes the automatic classification of the growth temperature of the seeding process,and the control system performs the temperature according to the classification result.After adjustment,it can meet the requirements of neck-growth temperature accuracy,successfully realize the automatic operation of the Czochralski crystal growth process,improve the control performance and automation level of such mainstream crystal growth equipment,and verify the proposed mean shifting.The applicability of the clustering algorithm reduced least squares support vector machine model.