摘要：

大数据分析方法能发现数据中存在的关系和规则,预测事物未来的发展趋势,从而提高决策的科学性。针对传统预测方法精度低、泛化性差的问题,提出基于智能支持向量机的大数据分析与预测方法。设计一种新的支持向量机模型参数选择准则,即模型残差概率密度函数逼近给定的高斯分布,并按照该准则采用混沌收缩粒子群优化算法确定模型参数,从而提高数据分类或回归处理的精度与泛化性。采用选矿生产过程现场数据进行实验, 结果验证了该方法的有效性,并表明其精度比LSSVM 方法更高。

关键词: 大数据, 支持向量机, 智能, 概率密度函数, 粒子群优化算法

Abstract:

It is important to study on methods of big data analysis that is used to find the relationship and rule between data to predict the future trend of things. This paper presents a big data analysis and forecasting method based on Support Vector Machine (SVM),which solves the problem that traditional prediction methods have low precision and poor generalization. The Probability Density Function(PDF) control based model parameters selection criterion is proposed to make the modeling error track a target Gaussian PDF. A contract Particle Swarm Optimization ( PSO) algorithm is adopted to tune the parameters. The proposed modeling approach is validated using the practical data,and the results show its efficiency. Compared with LSSVM,the accuracy of the proposed method is higher.

Key words: big data, Support Vector Machine(SVM), intelligent, Probability Density Function(PDF), Particle Swarm Optimization(PSO) algorithm

中图分类号: 

• TP391. 9