学业成绩的影响因素与预测模型研究

随着教育信息化、智慧化校园的建设，大量的学习和行为数据被记录和保存，利用这些教育数据探究学生成绩的影响因素，并对成绩进行预测成为教育数据挖掘领域（EDM）的重要研究问题。本文的主要目标是识别学业预警群体，通过加入新的影响因素变量和应对类别不平衡问题，从而提高模型整体的预测精度。为此，我们基于南方科技大学的真实学生数据集展开研究。首先我们考虑了数据集的背景与特点，从学习环境（线上：在线学习，线下：校园宿舍）角度，挖掘影响学生成绩的重要因素。同时为了解决类别不平衡问题所带来的模型预测效果不理想，引入基于K均值聚类的合成少数类过采样技术（KSMOTE）。最后，将影响因素作为特征变量加入到预测模型中以进一步检验新变量对模型的预测效果和解释能力。

结果表明，对于线下的学习环境而言，宿舍环境是个重要的影响因素。使用固定效应回归模型，研究表明宿舍距离图书馆越近，学生的成绩越高；通过中介效应分析发现，宿舍距离越近，学生去图书馆的次数越多，使用图书馆的讨论间越频繁，成绩往往越高；通过异质性分析发现，男性受到的“距离”影响显著的大于女性受到的影响，这表明男生可能更容易受到不利环境的影响。对于线上的学习环境而言，使用双重差分法（DID），我们发现男性在需要更多自控能力要求的线上学习环境下，与女性的成绩差距进一步拉大；使用无条件分位数回归模型（UQR），我们发现性别差距既在不同分数段的学生有所不同，还在不同学期有所不同。因此本文得到了“距离”，“性别”这两个影响成绩的变量。使用KSMOTE模型处理了类别不平衡问题后，成绩预测模型的效果得到了显著提高；进一步，在加入了“距离”、“性别”变量后，模型的预测效果进一步得到了改善。

[1] 肖巍, 倪传斌, 李锐. 国外基于数据挖掘的学习预警研究[J]. 中国远程教育, 2018: 70-78.
[2] 李宇帆, 张会福, 刘上力, 等. 教育数据挖掘研究进展[J]. 计算机工程与应用, 2019, 55:15-23.
[3] 周庆, 牟超, 杨丹. 教育数据挖掘研究进展综述[J]. 软件学报, 2015, 26: 3026-3042.
[4] ROMERO C, VENTURA S. Educational data mining: a review of the state of the art[J]. IEEETransactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews), 2010, 40(6): 601-618.
[5] 徐蕾, 聂峰英. 教育大数据在教育管理应用中的研究现状与技术路径[J]. 江苏高教, 2020:69-73.
[6] MORENO-MARCOS P M, MUÑOZ-MERINO P J, ALARIO-HOYOS C, et al. Analysingthe predictive power for anticipating assignment grades in a massive open online course[J].Behaviour & Information Technology, 2018, 37(10-11): 1021-1036.
[7] MARBOUTI F, DIEFES-DUX H A, MADHAVAN K. Models for early prediction of at-riskstudents in a course using standards-based grading[J]. Computers & Education, 2016, 103:1-15.
[8] DU X, YANG J, SHELTON B E, et al. A systematic meta-review and analysis of learninganalytics research[J]. Behaviour & information technology, 2021, 40(1): 49-62.
[9] GARDNER J, BROOKS C. Student success prediction in moocs[J]. User Modeling and UserAdapted Interaction, 2018, 28(2): 127-203.
[10] LUO J, SOROUR S E, GODA K, et al. Predicting student grade based on free-style commentsusing word2vec and ann by considering prediction results obtained in consecutive lessons.[J].International Educational Data Mining Society, 2015.
[11] JISHAN S T, RASHU R I, HAQUE N, et al. Improving accuracy of students’final grade prediction model using optimal equal width binning and synthetic minority over-sampling technique[J]. Decision Analytics, 2015, 2(1): 1-25.
[12] PÉREZ-LEMONCHE Á, MARTÍNEZ-MUÑOZ G, PULIDO-CAÑABATE E. Analysing eventtransitions to discover student roles and predict grades in moocs[C]//International Conferenceon Artificial Neural Networks. Springer, 2017: 224-232.
[13] HELLAS A, IHANTOLA P, PETERSEN A, et al. Predicting academic performance: a systematic literature review[C]//Proceedings companion of the 23rd annual ACM conference oninnovation and technology in computer science education. 2018: 175-199.
[14] PUNLUMJEAK W, RUGTANOM S, JANTARAT S, et al. Improving classification of imbalanced student dataset using ensemble method of voting, bagging, and adaboost with undersampling technique[M]//IT Convergence and Security 2017. Springer, 2018: 27-34.
[15] LAST F, DOUZAS G, BACAO F. Oversampling for imbalanced learning based on k-means andsmote[J]. arXiv preprint arXiv:1711.00837, 2017.
[16] RENDON E, ALEJO R, CASTORENA C, et al. Data sampling methods to deal with the bigdata multi-class imbalance problem[J]. Applied Sciences, 2020, 10(4): 1276.
[17] 姚丽, 崔超然, 马乐乐, 等. 基于校园上网行为感知的学生成绩预测方法[J]. 计算机研究与发展: 1-14.
[18] XU X, WANG J, PENG H, et al. Prediction of academic performance associated with internetusage behaviors using machine learning algorithms[J]. Computers in Human Behavior, 2019,98: 166-173.
[19] ASIF R, MERCERON A, ALI S A, et al. Analyzing undergraduate students’ performance usingeducational data mining[J]. Computers & Education, 2017, 113: 177-194.
[20] CAO Y, GAO J, LIAN D, et al. Orderliness predicts academic performance: behavioural analysison campus lifestyle[J]. Journal of The Royal Society Interface, 2018, 15(146): 20180210.
[21] 申航杰, 琚生根, 孙界平. 基于模糊聚类和支持向量回归的成绩预测[J]. 华东师范大学学报 ( 自然科学版 ), 2019: 66-73+84.
[22] 孟伟东. 数据挖掘在教育大数据中的应用探微[J]. 公关世界, 2021: 139-140.
[23] CHUAN Y Y, HUSAIN W, SHAHIRI A M. An exploratory study on students’performanceclassification using hybrid of decision tree and naïve bayes approaches[C]//International Conference on Advances in Information and Communication Technology. Springer, 2016: 142-152.
[24] ALBOANEEN D, ALMELIHI M, ALSUBAIE R, et al. Development of a web-based predictionsystem for students’academic performance[J]. Data, 2022, 7(2): 21.
[25] NATEK S, ZWILLING M. Student data mining solution–knowledge management system related to higher education institutions[J]. Expert systems with applications, 2014, 41(14): 6400-6407.
[26] 蒋卓轩, 张岩, 李晓明. 基于 mooc 数据的学习行为分析与预测[J]. 计算机研究与发展,2015, 52: 614-628.
[27] 常李艳, 刘婧, 黄崇. mooc 辍学研究：近 20 年文献的系统分析[J]. 图书馆论坛, 2021, 41:38-51.
[28] HAMSA H, INDIRADEVI S, KIZHAKKETHOTTAM J J. Student academic performanceprediction model using decision tree and fuzzy genetic algorithm[J]. Procedia Technology,2016, 25: 326-332.
[29] SAA A A, et al. Educational data mining & students’performance prediction[J]. InternationalJournal of Advanced Computer Science and Applications, 2016, 7(5): 212-220.
[30] 张欣欣, 郭纯, 郭真, 等. 结合灰度预测特征与 cnns 的信息服务体育成绩预测[J]. 沈阳工业大学学报, 2020, 42: 432-436.
[31] KEMPER L, VORHOFF G, WIGGER B U. Predicting student dropout: A machine learningapproach[J]. European Journal of Higher Education, 2020, 10(1): 28-47.
[32] BEAULAC C, ROSENTHAL J S. Predicting university students’academic success and majorusing random forests[J]. Research in Higher Education, 2019, 60(7): 1048-1064.
[33] HU Y H, LO C L, SHIH S P. Developing early warning systems to predict students’onlinelearning performance[J]. Computers in Human Behavior, 2014, 36: 469-478.
[34] LECUN Y, BENGIO Y, HINTON G. Deep learning[J]. nature, 2015, 521(7553): 436-444.
[35] DU X, YANG J, HUNG J L. An integrated framework based on latent variational autoencoderfor providing early warning of at-risk students[J]. IEEE Access, 2020, 8: 10110-10122.
[36] 杜晓明, 葛世伦, 王念新. 基于 cnn ＿ lstm 混合神经网络模型的学业预测[J]. 现代教育技术, 2021, 31: 69-76.
[37] 靳现凯, 宋威. 基于 dnn 的大学生学业成绩预测方法研究——以北京市某高校电子信息类专业为例[J]. 北方工业大学学报, 2021, 33: 134-140.
[38] QIU L, LIU Y, HU Q, et al. Student dropout prediction in massive open online courses byconvolutional neural networks[J]. Soft Computing, 2019, 23(20): 10287-10301.
[39] HANG-JIE S, SHENG-GEN J, JIE-PING S. Performance prediction based on fuzzy clusteringand support vector regression[J]. Journal of East China Normal University (Natural Science),2019, 2019(5): 66.
[40] 赵磊, 邓彤, 吴卓平. 基于数据挖掘的 mooc 学习者学业成绩预测与群体特征分析[J]. 重庆高教研究, 2021, 9: 95-105.
[41] 李梦莹, 王晓东, 阮书岚, 等. 基于双路注意力机制的学生成绩预测模型[J]. 计算机研究与发展, 2020, 57: 1729-1740.
[42] 陈辉, 白骏, 殷传涛, 等. 基于学习行为的 mooc 用户持续学习预测框架[J]. 北京航空航天大学学报: 1-10.
[43] OKUBO F, YAMASHITA T, SHIMADA A, et al. A neural network approach for students’ performance prediction[C]//Proceedings of the seventh international learning analytics & knowledge conference. 2017: 598-599.
[44] BRAHIM G B. Predicting student performance from online engagement activities using novelstatistical features[J]. Arabian Journal for Science and Engineering, 2022: 1-19.
[45] PERNA L, RUBY A, BORUCH R, et al. The life cycle of a million mooc users[C]//MOOCResearch Initiative Conference: volume 5. 2013.
[46] LÓPEZ V, FERNÁNDEZ A, GARCÍA S, et al. An insight into classification with imbalanceddata: Empirical results and current trends on using data intrinsic characteristics[J]. Informationsciences, 2013, 250: 113-141.
[47] 杨坤融, 熊余, 张健, 等. 面向长短期混合数据的 mooc 辍学预测策略研究[J]. 计算机工程与应用: 1-11.
[48] THAI-NGHE N, BUSCHE A, SCHMIDT-THIEME L. Improving academic performance prediction by dealing with class imbalance[C]//2009 Ninth international conference on intelligentsystems design and applications. IEEE, 2009: 878-883.
[49] 范逸洲, 刘敏, 欧阳嘉煜, 等. mooc 中学习者流失问题的预测分析——基于 24 篇中英文文献的综述[J]. 中国远程教育, 2018: 5-14+79.
[50] CHETTY R, FRIEDMAN J N, HILGER N, et al. How does your kindergarten classroom affectyour earnings? evidence from project star[J]. The Quarterly journal of economics, 2011, 126(4): 1593-1660.
[51] CHETTY R, FRIEDMAN J N, ROCKOFF J E. Measuring the impacts of teachers ii: Teachervalue-added and student outcomes in adulthood[J]. American economic review, 2014, 104(9):2633-79.
[52] SACERDOTE B. How large are the effects from changes in family environment? a study ofkorean american adoptees[J]. The Quarterly Journal of Economics, 2007, 122(1): 119-157.
[53] RAMLI A, ZAIN R M, CAMPUS C, et al. The impact of facilities on students’academicachievement[J]. Science International, 2018, 30(2): 299-311.
[54] GREENWALD R, HEDGES L V, LAINE R D. The effect of school resources on studentachievement[J]. Review of educational research, 1996, 66(3): 361-396.
[55] BURGESS B, KAYA N. Gender differences in student attitude for seating layout in collegeclassrooms[J]. College Student Journal, 2007, 41(4): 940-947.
[56] YANG Z, BECERIK-GERBER B, MINO L. A study on student perceptions of higher educationclassrooms: Impact of classroom attributes on student satisfaction and performance[J]. Buildingand Environment, 2013, 70: 171-188.
[57] DE ARAUJO P, MURRAY J. Estimating the effects of dormitory living on student performance[J]. Available at SSRN 1555892, 2010.
[58] DE ARAUJO P, MURRAY J. Channels for improved performance from living on campus.[J].American Journal of Business Education, 2010, 3(12): 57-64.
[59] WANG S, MAO Y. The effect of boarding on campus on left-behind children’s sense ofschool belonging and academic achievement: Chinese evidence from propensity score matchinganalysis[J]. Asia Pacific Journal of Education, 2018, 38(3): 378-393.
[60] WALKER A, MURRAY S. The relationship between anxiety, homesickess, academic performance, and social adjustment in college students[J]. FSU Journal of Behavioral Sciences, 2009,13(1): 4.
[61] CURTO V E, FRYER JR R G. The potential of urban boarding schools for the poor: Evidencefrom seed[J]. Journal of Labor Economics, 2014, 32(1): 65-93.
[62] SACERDOTE B. Peer effects with random assignment: Results for dartmouth roommates[J].The Quarterly journal of economics, 2001, 116(2): 681-704.
[63] BLOOM N, LEMOS R, SADUN R, et al. Does management matter in schools?[J]. The Economic Journal, 2015, 125(584): 647-674.
[64] ARAUJO M C, CARNEIRO P, CRUZ-AGUAYO Y, et al. Teacher quality and learning outcomesin kindergarten[J]. The Quarterly Journal of Economics, 2016, 131(3): 1415-1453.
[65] LUDWIG J, DUNCAN G J, GENNETIAN L A, et al. Long-term neighborhood effects on lowincome families: Evidence from moving to opportunity[J]. American economic review, 2013,103(3): 226-31.
[66] CHETTY R, HENDREN N, KATZ L F. The effects of exposure to better neighborhoods on children: New evidence from the moving to opportunity experiment[J/OL]. American EconomicReview, 2016, 106(4): 855-902. https://www.aeaweb.org/articles?id=10.1257/aer.20150572.
[67] BETTINGER E P, FOX L, LOEB S, et al. Virtual classrooms: How online college coursesaffect student success[J]. American Economic Review, 2017, 107(9): 2855-75.
[68] XU D, JAGGARS S S. Performance gaps between online and face-to-face courses: Differencesacross types of students and academic subject areas[J]. The Journal of Higher Education, 2014,85(5): 633-659.
[69] JAGGARS S, BAILEY T R. Effectiveness of fully online courses for college students: Responseto a department of education meta-analysis[J]. 2010.
[70] CASTRO M D B, TUMIBAY G M. A literature review: efficacy of online learning courses forhigher education institution using meta-analysis[J]. Education and Information Technologies,2021, 26(2): 1367-1385.
[71] FIGLIO D, RUSH M, YIN L. Is it live or is it internet? experimental estimates of the effects ofonline instruction on student learning[J]. Journal of Labor Economics, 2013, 31(4): 763-784.
[72] HUBER B, TRÜCK J, EICH G, et al. Febrile seizures in children during the influenza a (h1n1)pandemic 2009/2010[J]. Klinische Pädiatrie, 2011, 223(07): 438-439.
[73] ESNARD A M, LAI B S, WYCZALKOWSKI C, et al. School vulnerability to disaster: examination of school closure, demographic, and exposure factors in hurricane ike’s wind swath[J].Natural Hazards, 2018, 90(2): 513-535.
[74] PAUDEL J, RYU H. Natural disasters and human capital accumulation: The case of nepal’searthquake[J]. Available at SSRN 2933795, 2017.
[75] STUART K L, PATTERSON L G, JOHNSTON D M, et al. Managing temporary school closuredue to environmental hazard: Lessons from new zealand[J]. Management in Education, 2013,27(1): 25-31.
[76] WONG K K, SHI J, GAO H, et al. Why is school closed today? unplanned k-12 school closuresin the united states, 2011–2013[J]. PLoS One, 2014, 9(12): e113755.
[77] BAKER M. Industrial actions in schools: strikes and student achievement[J]. Canadian Journalof Economics/Revue canadienne d’économique, 2013, 46(3): 1014-1036.
[78] GOODMAN J. Flaking out: Student absences and snow days as disruptions of instructionaltime[R]. National Bureau of Economic Research, 2014.
[79] PISCHKE J S. The impact of length of the school year on student performance and earnings:Evidence from the german short school years[J]. The Economic Journal, 2007, 117(523): 1216-1242.
[80] MALDONADO J, DE WITTE K. The effect of school closures on standardised student test[J].British Educational Research Journal, 2020.
[81] UNESCO. Education: From disruption to recovery[EB/OL]. 2022. https://en.unesco.org/covid19/educationresponse.
[82] ALMAIAH M A, AL-KHASAWNEH A, ALTHUNIBAT A. Exploring the critical challengesand factors influencing the e-learning system usage during covid-19 pandemic[J]. Educationand information technologies, 2020, 25(6): 5261-5280.
[83] COPELAND W E, MCGINNIS E, BAI Y, et al. Impact of covid-19 pandemic on college student mental health and wellness[J]. Journal of the American Academy of Child & AdolescentPsychiatry, 2021, 60(1): 134-141.
[84] HOYT L T, COHEN A K, DULL B, et al. “constant stress has become the new normal”:stress and anxiety inequalities among us college students in the time of covid-19[J]. Journal ofAdolescent Health, 2021, 68(2): 270-276.
[85] WANG C, ZHAO H, ZHANG H. Chinese college students have higher anxiety in new semesterof online learning during covid-19: a machine learning approach[J]. Frontiers in psychology,2020: 3465.
[86] WANG C, ZHAO H. The impact of covid-19 on anxiety in chinese university students[J]. Frontiers in psychology, 2020, 11: 1168.
[87] CUNHA J, SILVA C, GUIMARÃES A, et al. No children should be left behind during covid19 pandemic: Description, potential reach, and participants’ perspectives of a project throughradio and letters to promote self-regulatory competences in elementary school[J]. Frontiers inpsychology, 2021, 12: 1308.
[88] REIMER D, SMITH E, ANDERSEN I G, et al. What happens when schools shut down? investigating inequality in students’reading behavior during covid-19 in denmark[J]. Researchin Social Stratification and Mobility, 2021, 71: 100568.
[89] TANG Y M, CHEN P C, LAW K M, et al. Comparative analysis of student’s live online learningreadiness during the coronavirus (covid-19) pandemic in the higher education sector[J]. Computers & Education, 2021, 168: 104211.
[90] PAROLIN Z, LEE E K. Large socio-economic, geographic and demographic disparities existin exposure to school closures[J]. Nature human behaviour, 2021, 5(4): 522-528.
[91] BONAL X, GONZÁLEZ S. The impact of lockdown on the learning gap: family and schooldivisions in times of crisis[J]. International Review of Education, 2020, 66(5): 635-655.
[92] KHLAIF Z N, SALHA S, KOURAICHI B. Emergency remote learning during covid-19 crisis:Students’engagement[J]. Education and information technologies, 2021, 26(6): 7033-7055.
[93] 罗长远, 司春晓. 在线教育会拉大不同家庭条件学生的差距吗?[J]. 财经研究, 2020, 46(11): 4-18.
[94] RODRIGUEZ-BESTEIRO S, TORNERO-AGUILERA J F, FERNÁNDEZ-LUCAS J, et al.Gender differences in the covid-19 pandemic risk perception, psychology and behaviors of spanish university students[J]. International Journal of Environmental Research and Public Health,2021, 18(8): 3908.
[95] ANDREW A, CATTAN S, COSTA DIAS M, et al. Inequalities in children’s experiences ofhome learning during the covid-19 lockdown in england[J]. Fiscal Studies, 2020, 41(3): 653-683.
[96] FISHER A N, RYAN M K. Gender inequalities during covid-19[J]. Group Processes & Intergroup Relations, 2021, 24(2): 237-245.
[97] HENNEKAM S, SHYMKO Y. Coping with the covid-19 crisis: Force majeure and genderperformativity[J]. Gender, Work & Organization, 2020, 27(5): 788-803.
[98] ROMERO C, VENTURA S. Data mining in education[J]. Wiley Interdisciplinary Reviews:Data Mining and Knowledge Discovery, 2013, 3(1): 12-27.
[99] LEITNER P, KHALIL M, EBNER M. Learning analytics in higher education—a literaturereview[J]. Learning analytics: Fundaments, applications, and trends, 2017: 1-23.
[100] RUBIN D B. Estimating causal effects of treatments in randomized and nonrandomized studies.[J]. Journal of educational Psychology, 1974, 66(5): 688.
[101] 温忠麟, 叶宝娟. 中介效应分析[J]. 心理科学进展, 2014, 22: 731-745.
[102] BARON R M, KENNY D A. The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations[J]. J. Pers. Soc. Psychol.,1986, 51(6): 1173-1182.
[103] KOENKER R, BASSETT JR G. Regression quantiles[J]. Econometrica: journal of the Econometric Society, 1978: 33-50.
[104] 朱平芳, 张征宇. 无条件分位数回归[J]. 统计研究, 2012, 29: 88-96.
[105] FIRPO S, FORTIN N M, LEMIEUX T. Unconditional quantile regressions[J]. Econometrica,2009, 77(3): 953-973.
[106] KREMER M, LEVY D. Peer effects and alcohol use among college students[J]. Journal ofEconomic perspectives, 2008, 22(3): 189-206.
[107] ZHANG X M, ZHU F. Group size and incentives to contribute: A natural experiment at chinesewikipedia[J]. American Economic Review, 2011, 101(4): 1601-15.
[108] 郭晔, 黄振, 姚若琪. 战略投资者选择与银行效率——来自城商行的经验证据[J]. 经济研究, 2020, 55: 181-197.
[109] CUI R, DING H, ZHU F. Gender inequality in research productivity during the COVID-19pandemic. manufacturing & service operations management[M]. 2021.
[110] LAVY V. Do gender stereotypes reduce girls’ or boys’ human capital outcomes? evidence froma natural experiment[J]. J. Public Econ., 2008, 92(10-11): 2083-2105.
[111] AZMAT G, CALSAMIGLIA C, IRIBERRI N. Gender differences in response to big stakes[J].J. Eur. Econ. Assoc., 2016, 14(6): 1372-1400.
[112] BRENØE A A, ZÖLITZ U. Exposure to more female peers widens the gender gap in STEMparticipation[J]. J. Labor Econ., 2020(706646): 000-000.
[113] MORIN L P. Do men and women respond differently to competition? evidence from a majoreducation reform[J]. J. Labor Econ., 2015, 33(2): 443-491.
[114] NIEDERLE M, VESTERLUND L. Do women shy away from competition? do men competetoo much?[J]. Q. J. Econ., 2007, 122(3): 1067-1101.
[115] 孙琳琳, 杨浩, 郑海涛. 土地确权对中国农户资本投资的影响——基于异质性农户模型的微观分析[J]. 经济研究, 2020, 55: 156-173.
[116] 温忠麟, 侯杰泰, 张雷. 调节效应与中介效应的比较和应用[J]. 心理学报, 2005: 268-274.
[117] CHEN Y, FANG H. The long-term consequences of china’s “later, longer, fewer”campaignin old age[J]. Journal of Development Economics, 2021, 151: 102664.
[118] NUNN N, QIAN N. The potato’s contribution to population and urbanization: evidence from ahistorical experiment[J]. The quarterly journal of economics, 2011, 126(2): 593-650.
[119] CUI R, DING H, ZHU F. Gender inequality in research productivity during the covid-19 pandemic[J]. Manufacturing & Service Operations Management, 2021.
[120] CHAWLA N V, BOWYER K W, HALL L O, et al. Smote: synthetic minority over-samplingtechnique[J]. Journal of artificial intelligence research, 2002, 16: 321-357.
[121] MACQUEEN J, et al. Some methods for classification and analysis of multivariate observations[C]//Proceedings of the fifth Berkeley symposium on mathematical statistics and probability:volume 1. Oakland, CA, USA, 1967: 281-297.
[122] BAEZA-YATES R, RIBEIRO-NETO B, et al. Modern information retrieval: volume 463[M].ACM press New York, 1999.
[123] DUDANI S A. The distance-weighted k-nearest-neighbor rule[J]. IEEE Transactions on Systems, Man, and Cybernetics, 1976(4): 325-327.
[124] QUINLAN J R. Induction of decision trees[J]. Machine learning, 1986, 1(1): 81-106.
[125] BREIMAN L. Random forests[J]. Machine learning, 2001, 45(1): 5-32.
[126] BOSER B E, GUYON I M, VAPNIK V N. A training algorithm for optimal margin classifiers[C]//Proceedings of the fifth annual workshop on Computational learning theory. 1992: 144-152.