摘要
研究了基于加权支持向量机的概率密度估计算法。现有算法只考虑采样时间或样本密度,导致概率密度结果误差较大。为提升估计精度,文中提出了一种改进的W-SVM算法,该算法同时考虑了采用时间和区域样本点,依次选择不同的加权惩罚系数,并对加权系数进行归一化处理,最后采用网格寻优法找出最合适的加权系数。仿真结果表明,所提改进加权支持向量的概率密度估计的均方误差远小于传统算法,即所提算法优于传统算法。
The probability density estimation algorithm based on W-SVM is studied. The existing algorithms only consider either the sampling time or the sample density,resulting in lager error probability density results. An improved W-SVM algorithm is proposed to improve the estimation precision by considering both the sample point of time and the use of area. Different types of penalty weighting coefficients are selected and normalized,and the most appropriate weighting factor is found by the grid optimization method. The simulation results show that the mean square error by the proposed improved probability weighted support vector density estimation is much smaller than that by traditional algorithms.
出处
《电子科技》
2014年第9期40-43,共4页
Electronic Science and Technology
基金
四川师范大学成都学院国家级大学生创新训练基金资助项目(201213672003)
