Parameter selection of support vector machine based on stepped-up chaos optimization algorithm
-
摘要: 针对支持向量机参数选择问题, 以惩罚系数、不敏感系数和RBF核函数中的宽度系数为优化变量, 采用Chebyshev映射代替Logistic映射产生初始混沌序列, 改变原有的搜索公式及增加3次载波, 提出了一种改进的加速混沌优化算法(ISCOA)。将该算法应用于人工数据集和实际数据集中, 并与常规的交叉验证法进行比较。试验结果表明: 在人工数据集中, 采用ISCOA在时间上缩短了至少23.43%, 精度上提高了至少6.31%;在实际数据集中, 预测值更接近实际值, 相对误差均控制在3.13%以下, 该算法具有较高的预测精度和寻优效果。Abstract: In order to analyze the parameter selection of support vector machine (SVM), penalty coefficient, insensitive coefficient and width coefficient in radial basis function (RBF) were used as optimization variables, the former searching formula was changed, and the third searching time was added. A new improved stepped-up chaos optimization algorithm (ISCOA) was proposed by adopting the Chebyshev mapping instead of Logistic mapping to form initial chaos serial. The new algorithm was used in artificial data set and real data set, and was compared with traditional cross validation method. Test result indicates that the running time is cut down at least 23.43%, and the precision improves at least 6.31% by using ISCOA in artificial data set. The predicted value is more close to real value, and the relative errors are controlled under 3.13% in real data set. So ISCOA has higher prediction precision and optimization effect.
-
表 1 试验结果比较
Table 1. Comparison of test results
指标 34个样本(δ=0.05) 67个样本(δ=0.10) CV算法 ISCOA CV算法 ISCOA 时间/s 223.56 139.42 321.58 246.25 MSE值 0.002 05 0.001 73 0.001 62 0.001 43 MRE值 1.357 60 1.035 80 0.815 70 0.764 20 
下载: 导出CSV
表 2 货物吞吐量比较
Table 2. Comparison of cargo throughputs
年份 实际值/104 t CV算法 ISCOA 预测值/104 t 相对误差 预测值/104 t 相对误差 2001 8 653 9 021.54 0.042 6 8 923.47 0.031 3 2002 10 638 10 660.49 0.002 1 10 428.26 0.019 7 2003 12 968 12 999.07 0.002 4 12 990.59 0.001 7 2004 15 836 15 688.92 0.009 3 15 951.23 0.007 3 2005 18 492 18 373.91 0.006 4 18 421.81 0.003 8
下载: 导出CSV
表 3 集装箱吞吐量比较
Table 3. Comparison of container throughputs
年份 实际值/104 TEU CV算法 ISCOA 预测值/104 TEU 相对误差 预测值/104 TEU 相对误差 2001 634.1 654.98 0.032 9 648.73 0.023 1 2002 861.4 841.40 0.023 2 873.85 0.014 5 2003 1 128.0 1 116.36 0.010 3 1 136.38 0.007 4 2004 1 455.0 1 448.18 0.004 7 1 450.52 0.003 1 2005 1 808.0 1 837.66 0.016 4 1 789.14 0.010 4
下载: 导出CSV
-
[1] LI U Huan-jun, WANG Yao-nan, LU Xiao-fen. A method to choose kernel function and its parameters for support vector machines[C]∥South China University of Technology. Proceedings of the Fourth International Conference on Machine Learning and Cybernetics. Guangzhou: Springer-Verlag, 2005: 4277-4280. [2] SMOLA AJ, SCH LKOPF B. Atutorial on support vecto rregression[J]. Statistics and Computing, 2004, 14 (3): 199-222. doi: 10.1023/B:STCO.0000035301.49549.88 [3] CHU W, KEERTHI S S. Newapproaches to support vector ordinal regression[C]∥ University of Bonn. Proceedings of the 22nd International Conference on Machine Learning. Bonn: ACM, 2005: 145-152. [4] KHAN MS, COULIBALY P. Application of support vector machine in lake water level prediction[J]. Journal of Hydrologic Engineering, 2006, 11 (3): 199-205. doi: 10.1061/(ASCE)1084-0699(2006)11:3(199) [5] 张旭东, 陈锋, 高隽, 等. 稀疏贝叶斯及其在时间序列预测中的应用[J]. 控制与决策, 2006, 21 (5): 585-588. doi: 10.3321/j.issn:1001-0920.2006.05.023ZHANG Xu-dong, CHEN Feng, GAO Jun, et al. Sparse Bayesian and its application to ti me series forecasting[J]. Conrol and Decision, 2006, 21 (5): 585-588. (in Chinese) doi: 10.3321/j.issn:1001-0920.2006.05.023 [6] WU C H, HO J M, LEE D T. Travel-time prediction with support vector regression[J]. IEEE Transactions on Intelligent Transportation Systems, 2004, 5 (4): 276-281. doi: 10.1109/TITS.2004.837813 [7] DONGJ X, KRZYZAK A, SUEN C Y. Fast SVM training algorithm with decomposition on very large data sets[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2005, 27 (4): 603-618. doi: 10.1109/TPAMI.2005.77 [8] 刘胜, 李妍妍. 自适应GA-SVM参数选择算法研究[J]. 哈尔滨工程大学学报, 2007, 28 (4): 398-402. doi: 10.3969/j.issn.1006-7043.2007.04.007LI USheng, LI Yan-yan. Parameter selection algorithm for support vector machines based on adaptive genetic algorithm[J]. Journal of Harbin Engineering University, 2007, 28 (4): 398-402. (in Chinese) doi: 10.3969/j.issn.1006-7043.2007.04.007 [9] CAWLEY G C, TALBOT N L C. Fast exact leave-one-out cross-validation of sparse least-squares support vector machines[J]. Neural Networks, 2004, 17 (10): 1467-1475. doi: 10.1016/j.neunet.2004.07.002 [10] 李昊, 胡云昌, 曹宏铎. 加速混沌优化方法及其应用[J]. 系统工程学报, 2002, 17 (1): 41-44. doi: 10.3969/j.issn.1000-5781.2002.01.008LI Ying, HU Yun-chang, CAO Hong-duo. Stepped-up chaos optimization algorithm and its application[J]. Journal of Systems Engineering, 2002, 17 (1): 41-44. (in Chinese) doi: 10.3969/j.issn.1000-5781.2002.01.008 [11] 李冬琴, 王呈方, 王丽铮. 基于单参数的Lagrangian支持向量回归算法及其应用[J]. 中国航海, 2007 (3): 73-76. doi: 10.3969/j.issn.1000-4653.2007.03.019LI Dong-qin, WANG Cheng-fang, WANG Li-zheng. A single-parameter Lagrangian support vector regression andits application[J]. Navigation of China, 2007 (3): 73-76. (in Chinese) doi: 10.3969/j.issn.1000-4653.2007.03.019 [12] 李冬琴, 王丽铮, 王呈方. 一种新的支持向量回归算法及其在集装箱吞吐量预测中的应用[J]. 水运工程, 2007 (5): 9-12. doi: 10.3969/j.issn.1002-4972.2007.05.003LI Dong-qin, WANG Li-zheng, WANG Cheng-fang. Prediction of container throughput using a single-parameter Lagrangian support vector regression[J]. Port and Waterway Engineering, 2007 (5): 9-12. (in Chinese) doi: 10.3969/j.issn.1002-4972.2007.05.003 -
点击查看大图
图(3) / 表(3)
计量
- 文章访问数: 464
- HTML全文浏览量: 65
- PDF下载量: 727
- 被引次数: 0
