【作者】 罗锦坤；

【导师】 元昌安；

【作者基本信息】 广西师范学院 ， 计算机应用技术， 2012， 硕士

【摘要】 近几年我国高校突发事件的频繁发生，给学校本身及整个社会造成重大影响。挖掘高校突发事件的诱发因素及结果因素之间的关联关系，对有效预防高校突发事件具有重要意义。然而传统的关联规则挖掘算法在挖掘过程中需要多次重复地扫描数据库，同时产生大量候选集，效率比较低。因此，针对高校突发事件数据集，如何建立一个高效的关联规则挖掘算法是当前的一个亟待解决的问题。针对传统关联规则挖掘算法存在的不足，本文将把GEP和复杂网络引入到关联规则挖掘当中，利用复杂网络构造高校突发事件复杂模型，并结合GEP的简单基因编码在全局范围内搜索最优解的能力来优化关联规则挖掘的原始数据，旨在减少关联规则挖掘的时间，从而弥补传统算法不足。本文主要工作如下：第一，本文针对高校突发事件特点，提取事件各个属性，根据事件属性划分为诱发因素和后果因素，然后进行数据清理，把数据整合到一个数据库。第二，根据高校突发事件特征，确定复杂网络的节点和边，构造高校突发事件网络模型。第三，在此基础上，引入基因表达式编程优化算法全局搜索最优解，提出GEP-ECD(Event Network Community Division AlgorithmBased on The GEP，GEP-ECD)社区划分算法，对复杂网络突发事件模型进行社区划分，输出社区表。第四，提出GCAR(Association Rules Algorithm Based on the GEPand Complex Network，GCAR)新算法，针对每个社区表进行局部关联规则挖掘，以置信度和支持度作为强关联规则的衡量指标，挖掘有效规则，这些规则将为预测高校突发事件提供依据。第五，本文最后结合实例利用这些有用的规则建立高校突发事件预测模型，实验表明，该模型取得较好的效果。更多还原

【Abstract】 University emergencies frequently occurred in the past few years,seriously impact on the schools themselves and the whole society. Miningthe potential association between the university emergencies predisposingfactors and result factors, it is great value to be effective in preventingfrom campus emergencies.However, traditional association rules mining algorithm needs toscan the database with several times in mining process, and produced alarge number of candidates set at the same time. Therefore, in connectionwith the set of campus emergencies, how to set up an efficient algorithmfor mining association rules, it is the current problems to be solved.In response to the shortcomings of traditional association rulemining algorithm, in this thesis, the GEP(Gene Expression Programming)and complex network will be introduced into association rule mining, andthe use of complex network to tectonic complex universities emergenciesmodel, combine with GEP simple genetic code and the ability of globallysearch for optimal solutions to optimize the original data for associationrules mining, in order to reduce the time of mining association rules, so asto make up for the lack of traditional algorithms. The main work of thisthesis as follow:Firstly,this thesis aims at the features of campus emergencies and theabstraction of the attributes of each emergency, according to event attributes, divide it into predisposing factors and result factors, and thenmake the data clean, Integrate data from different data sources into adatabase.Secondly, according to the campus emergencies characteristics,determine the nodes and edges of complex network, use complex networkmodel to construct universities emergencies.Thirdly, on this basis, introduce gene expression programmingalgorithm for global search optimal solution, propose communitypartitioning algorithm of GEP-ECD, divide complex network model ofemergency into communities and output each community table.Fourthly, propose GCAR algorithm for mining each local associationrule community table, take the Confidence and Support as the evaluationcriterion for mining the effective rules, these rules will provide the basisfor predicting campus emergencies.Fifthly, combine instance with using these useful rules to establishuniversities emergency prediction model, and the experiments show thatthis model achieve the better results.更多还原