微阵列高维特征选择的多策略混合人工蜂群算法

doi:10.16182/j.issn1004731x.joss.21-1188

摘要/Abstract

摘要：

传统的特征选择方法对于高维微阵列具有较大的局限性，难以准确高效地提出最佳特征子集。针对该问题，提出了基于wrapper的多策略混合人工蜂群算法，该算法混合了混沌反向学习策略、精英引导策略、Mantegna Lévy分布策略，分别在雇佣蜂与观察蜂阶段提出了两种新的搜索策略。针对于微阵列高维特征选择问题，提出新的平衡模型性能最优与特征子集规模最小化目标函数。实验结果表明：该算法能够达到较高的分类准确率，可在一定程度上取得特征子集规模最小化的目标，且优于GABC等改进算法与樽海鞘群等六种新型智能算法。

关键词: 人工蜂群算法, 高维特征选择, 混沌反向学习策略, 精英引导策略, Mantegna Lévy分布

Abstract:

Traditional feature selection approaches have major limitations for high-dimensional microarrays, and it is difficult to accurately and efficiently propose the best feature subset. To address this problem, a multi-strategy hybrid artificial bee colony (ABC) algorithm based on wrapper is proposed, which mixes chaotic opposition-based learning strategy, elite guidance strategy, and Mantegna Lévy distribution strategy, and proposes two new search strategies in the employed and onlooker bee phases respectively. A new objective function is proposed for the microarray high-dimensional feature selection problem, which balances the optimal performance of the model with the minimization of the feature subset size. Experimental results show that the algorithm is able to achieve high classification accuracy while still satisfying the feature subset size minimization objective to some extent. Moreover, it outperforms improved algorithms such as GABC and six new intelligent algorithms such as the salp swarm algorithm.

Key words: artificial bee colony(ABC) algorithm, high-dimensional feature selection, chaotic opposition-based learning, elite guidance strategy, Mantegna Lévy distribution

中图分类号:

TP181

秦传东, 李宝胜, 韩宝乐. 微阵列高维特征选择的多策略混合人工蜂群算法[J]. 系统仿真学报, 2023, 35(3): 515-524.

Chuandong Qin, Baosheng Li, Baole Han. Multi-strategy Hybrid ABC for Microarray High-Dimensional Feature Selection[J]. Journal of System Simulation, 2023, 35(3): 515-524.

图/表 8

图1

图2

表1

表2

多种算法参数设置

算法	参数
ABC及改进	初始种群SN=20；最大函数评估次数MNFE=2 000；开采次数limit=50
SSA	初始种群SN=20；最大函数评估次数MNFE=2 000；控制参数 $c 1 = 2 e x p [- (4 L / i t e r m a x)]$
ASO	初始种群SN=20；最大函数评估次数MNFE=2 000；深度权重 $α = 50$ ；乘数权重 $β = 0.2$
PFA	初始种群SN=20；最大函数评估次数MNFE=2 000； $α, β ∈ [1, 2]$ ； $u ∈ [- 1, 1]$
BOA	初始种群SN=20；最大函数评估次数MNFE=2 000；感官模态 $c = 0.01$ ；幂指数 $a = 0.1 + 0.2 t / i t e r m a x$ ；开关概率 $P = 0.8$
GWO	初始种群SN=20；最大函数评估次数MNFE=2 000；收敛因子 $a = 2 (1 - i t e r / i t e r m a x)$

表2

表3

图3

表4

图4

参考文献 21

1	Sun L, Zhang X, Qian Y, et al. Feature selection Using Neighborhood Entropy-Based Uncertainty Measures for Gene Expression Data Classification[J]. Information Sciences (S0020-0255), 2019, 502: 18-41.
2	Nguyen B H, Xue B, Zhang M. A Survey on Swarm Intelligence Approaches to Feature Selection in Data Mining[J]. Swarm and Evolutionary Computation (S2210-6502), 2020, 54: 100663.
3	Meyer P E, Schretter C, Bontempi G. Information-Theoretic Feature Selection in Microarray Data Using Variable Complementarity[J]. IEEE Journal of Selected Topics in Signal Processing (S1932-4553), 2008, 2(3): 261-274.
4	Guyon I, Weston J, Barnhill S, et al. Gene Selection for Cancer Classification Using Support Vector Machines[J]. Machine Learning (S1573-0565), 2002, 46(1): 389-422.
5	Lee S J, Xu Z, Li T, et al. A Novel Bagging C4. 5 Algorithm Based on Wrapper Feature Selection for Supporting Wise Clinical Decision Making[J]. Journal of Biomedical Informatics (S1532-0464), 2018, 78: 144-155.
6	李炜, 巢秀琴. 改进的粒子群算法优化的特征选择方法[J]. 计算机科学与探索, 2019, 13(6): 990-1004.
	Li Wei, Chao Xiuqin. Improved Particle Swarm Optimization Method for Feature Selection[J]. Frontiers of Computer Science and Technology, 2019, 13(6): 990-1004.
7	Arslan S, Ozturk C. Multi Hive Artificial Bee Colony Programming for High Dimensional Symbolic Regression with Feature Selection[J]. Applied Soft Computing (S1568-4946), 2019, 78: 515-527.
8	Wang X, Zhang Y, Sun X, et al. Multi-objective Feature Selection Based on Artificial Bee Colony: An Acceleration Approach with Variable Sample Size[J]. Applied Soft Computing (S1568-4946), 2020, 88: 106041.
9	Rao H, Shi X, Rodrigue A K, et al. Feature Selection Based on Artificial Bee Colony and Gradient Boosting Decision Tree[J]. Applied Soft Computing (S1568-4946), 2019, 74: 634-642.
10	张戈, 王建林. 基于混合ABC和CRO的高维特征选择方法[J]. 计算机工程与应用, 2019, 55(11): 93-101.
	Zhang Ge, Wang Jianlin. Feature Selection of High-dimensional Data Based on ABC and CRO Algorithm[J]. Computer Engineering and Applications, 2019, 55(11): 93-101.
11	匡芳君, 金忠, 徐蔚鸿, 等. Tent混沌人工蜂群与粒子群混合算法[J]. 控制与决策, 2015, 30(5): 839-847.
	Kuang Fangjun, Jin Zhong, Xu Weihong, et al. Hybridization Algorithm of Tent Chaos Artificial Bee Colony and Particle Swarm Optimization[J]. Control and Decision, 2015, 30(5): 839-847.
12	高卫峰, 刘三阳, 焦合华, 等. 引入人工蜂群搜索算子的粒子群算法[J]. 控制与决策, 2012, 27(6): 833-838.
	Gao Weifeng, Liu Sanyang, Jiao Hehua, et al. Particle Swarm Optimization with Search Operator of Artificial Bee Colony Algorithm[J]. Control and Decision, 2012, 27(6): 833-838.
13	Gao W, Liu S, Huang L. Enhancing Artificial Bee Colony Algorithm Using More Information-Based Search Equations[J]. Information Sciences (S0020-0255), 2014, 270: 112-133.
14	Abualigah L, Shehab M, Alshinwan M, et al. Salp Swarm Algorithm: A Comprehensive Survey[J]. Neural Computing and Applications (S1433-3058), 2020, 32(15): 11195-11215.
15	Zhao W, Wang L, Zhang Z. Atom Search Optimization and Iits Application to Solve a Hydrogeologic Parameter Estimation Problem[J]. Knowledge-Based Systems (S0950-7051), 2019, 163: 283-304.
16	Yapici H, Cetinkaya N. A New Meta-heuristic Optimizer: Pathfinder Algorithm[J]. Applied Soft Computing (S1568-4946), 2019, 78: 545-568.
17	Arora S, Singh S. Butterfly Optimization Algorithm: A Novel Approach for Global Optimization[J]. Soft Computing (S1433-7479), 2019, 23(3): 715-734.
18	Mirjalili S, Mirjalili S M, Lewis A. Grey Wolf Optimizer[J]. Advances in Engineering Software (S0965-9978), 2014, 69: 46-61.
19	Zhu G, Kwong S. Gbest-Guided Artificial Bee Colony Algorithm for Numerical Function Optimization[J]. Applied Mathematics and Computation (S0096-3003), 2010, 217(7): 3166-3173.
20	Gao W, Liu S, Huang L. A Novel Artificial Bee Colony Algorithm Based on Modified Search Equation and Orthogonal Learning[J]. IEEE Transactions on Cybernetics (S2168-2275), 2013, 43(3): 1011-1024.
21	Luo J, Wang Q, Xiao X. A Modified Artificial Bee Colony Algorithm Based on Converge-Onlookers Approach for Global Optimization[J]. Applied Mathematics and Computation (S0096-3003), 2013, 219(20): 10253-10262.

数据集	实例数	特征数	类别数
GLI	85	22 283	2
GLA-BRA	180	49 151	4
CLL-SUB	111	11 340	3
TOX	171	5 748	4
SMK-CAN	187	19 993	2

数据集	指标	ABC		GABC		CABC		EABC		COABC		MSHABC-FS
数据集	指标	Mean	Std	Mean	Std	Mean	Std	Mean	Std	Mean	Std	Mean	Std
GLI	1	10 959	106.1	11 100	142.1	10 580	70.8	10 969	57.23	10 660	71.8	734.4	222.2
	2	0.198	0.016	0.206	0.019	0.214	0.011	0.228	0.012	0.023	0.212	0.013	0.003
	3	98.24	0.028	87.65	0.065	94.71	0.019	93.53	0.018	95.29	0.037	100	0
	4	25.77	0.075	43.29	0.336	23.55	0.095	296.22	1.299	23.50	0.115	31.10	0.27
GLA-BRA	1	24 367.6	159.7	24 483.8	125.1	23 687.8	186.2	24 362.2	153.8	23 884.8	85.45	1 628	705.8
	2	0.292	0.001	0.302	0.009	0.304	0.010	0.312 9	0.008	0.302	0.008	0.123	0.014
	3	83.33	0	78.33	0.030	80.56	0.019	80.00	0.012	81.11	0.012	82.22	0.025
	4	102.32	0.31	142.49	0.85	93.81	0.57	688.39	1.907	94.58	0.458	79.72	3.11
CLL-SUB	1	5 545.6	37.5	5 620.6	57.46	5 301.2	50.2	5 569.2	49.33	5 353.9	44.03	666.9	424.7
	2	0.327	0.001	0.328	0.002	0.319	0.013	0.348	0.019	0.329	0.019	0.140	0.032
	3	77.27	0	65.91	0.096	77.27	0.021	74.09	0.031	75.91	0.307	80.91	0.036
	4	19.03	0.099	28.04	0.105	17.23	0.109	156.57	1.093	17.45	0.128	20.02	0.433
TOX	1	2 854.7	36.7	2 866.7	32.43	2 658.6	56.68	2 833	46.41	2 691	26.85	284.4	136.45
	2	0.328	0.009	0.341	0.018	0.333	0.022	0.363	0.023	0.310	0.032	0.126	0.011
	3	77.65	0.015	59.12	0.058	74.71	0.034	71.76	0.037	78.82	0.051	82.65	0.017
	4	17.66	0.227	22.62	0.152	16.101	0.240	88.17	0.934	15.81	0.068	14.75	0.314
SMK-CAN	1	9 889.5	102.4	9 401.4	58.4	9 569.7	58.2	9 871.8	62.1	9 550.4	87.0	857.1	317.33
	2	0.327	0.007	0.323	0.016	0.323	0.014	0.332	0.008	0.329	0.006	0.108	0.012
	3	77.57	0.013	76.76	0.024	77.30	0.023	76.76	0.015	76.22	0.012	85.14	0.019
	4	48.27	0.147	44.32	0.196	43.82	0.165	287.09	1.029	43.93	0.249	37.19	0.862

数据集	指标	SSA		ASO		PFA		BOA		GWO		MSHABC-FS
数据集	指标	Mean	std	Mean	std	Mean	std	Mean	std	Mean	std	Mean	std
GLI	1	10 992	50.1	10 677	120.4	11 013	115.2	3 810.9	754.7	2 658.1	336.1	734.4	222.2
	2	0.212	0.017	0.205	0.017	0.203	0.017	0.076	0.013	0.046	0.006	0.013	0.003
	3	95.88	0.03	96.47	0.03	97.65	0.03	98.24	0.028	100	0	100	0
	4	23.43	0.07	30.73	0.39	26.26	0.08	18.04	0.13	18.05	0.63	31.10	0.27
GLA-BRA	1	24 347.6	121.9	23 875.2	227.9	24 396.2	132.1	6 577.2	1 382.8	6 184	524.5	1 628	705.8
	2	0.306	0.007	0.299	0.008	0.293	0.001	0.164	0.006	0.141	0.009	0.123	0.014
	3	81.11	0.012	81.67	0.015	83.33	0	81.67	0.015	85.00	0.015	82.22	0.025
	4	94.04	0.26	98.53	3.96	100.37	0.37	65.47	0.393	52.61	1.61	79.72	3.11
CLL-SUB	1	5 557.8	73.2	5 436.5	73.4	5 549.1	33.32	1 508.2	396.5	1 474.1	174.5	666.9	424.7
	2	0.339	0.013	0.324	0.003	0.327	0.001	0.211	0.019	0.159	0.015	0.140	0.032
	3	75.45	0.023	77.27	0	77.27	0	74.09	0.037	82.27	0.026	80.91	0.036
	4	17.19	0.120	21.18	0.597	18.58	0.047	13.04	0.082	13.34	0.345	20.02	0.433
TOX	1	2 826.3	33.4	2 769.8	54.99	2 838.3	20.29	705.4	113.6	965.6	117.8	284.4	136.45
	2	0.344	0.016	0.302	0.015	0.337	0.012	0.201	0.026	0.135	0.025	0.126	0.011
	3	74.71	0.030	80.88	0.025	75.88	0.019	75.00	0.044	88.53	0.039	82.65	0.017
	4	16.24	0.302	20.16	0.432	17.11	0.148	12.71	0.120	12.26	0.457	14.75	0.314
SMK-CAN	1	9 867.2	34.24	9 624	66.2	9 898	71.9	2 367.8	208.9	2 923.9	342.7	857.1	317.33
	2	0.347	0.014	0.322	0.015	0.326	0.014	0.192	0.007	0.123	0.020	0.108	0.012
	3	74.32	0.014	77.57	0.026	77.84	0.025	76.22	0.011	89.19	0.031	85.14	0.019
	4	44.33	0.010	42.14	1.601	47.44	0.054	31.24	0.217	27.20	1.395	37.19	0.862

[1]	郭旭东, 鲁正平, 蒋琴芬, 王殊轶, 崔海坡. 胃肠道给药胶囊的方位跟踪求解算法[J]. 系统仿真学报, 2018, 30(6): 2288-2294.
[2]	崔秋丽. 基于人工蜂群算法的微网并网优化运行[J]. 系统仿真学报, 2018, 30(4): 1551-1559.
[3]	杨小健, 董毅伟. 基于反向学习的自适应快速人工蜂群算法[J]. 系统仿真学报, 2016, 28(11): 2684-2692.
[4]	张明, 田娜, 纪志成. 基于适应值欧式距离比的均衡蜂群算法[J]. 系统仿真学报, 2015, 27(5): 980-989.