一种改进的遗传约简算法及其在查找“黑广播”中的应用

发布时间：2019-02-16 02:40

【摘要】：遗传约简算法是粗糙集约简算法之一。约简以保持信息系统分类能力不变为前提,删除冗余属性,获得泛化能力更强的决策规则或分类规则。属性约简在避免“维数灾难”,提高算法运行效率等方面具有重要意义。非法广播俗称“黑广播”,会影响无线电秩序,甚至干扰航空频段的正常使用。宣传虚假、低俗广告的广播是常见“黑广播”之一。智能化监测是查找“黑广播”的有效措施之一。不进行整点报时,以语音为主是大多数“黑广播”的重要特征。本文核心内容如下:(1)提出了整点报时识别算法。经过预处理和特征提取,建立整点报时和不报时调频广播信号的隐马尔可夫模型,使用实测数据验证本文方法识别整点报时调频广播信号的有效性。结合常规监测指标构建查找“黑广播”信息系统。(2)提出了基于等价关系信息系统的变长度染色体遗传约简算法。以变长度染色体遗传算法为基本框架,基于简化二进制差别矩阵的行列重要度构造适应度函数。引入调整函数使进化过程自适应地进行交叉和变异操作,对个体选择性地进行遗传操作。该遗传约简算法适用于相容决策表和不相容决策表。本文对多个UCI数据集进行属性约简,通过实验对比验证了本文算法的有效性和优越性。(3)提出了基于优势—等价关系信息系统的变长度染色体遗传约简算法。本文基于优势—等价关系信息系统的正域设计了适应度函数,实现了对条件属性值具有偏序关系,决策属性值是等价关系的决策表的属性约简。实验结果表明,本文算法与传统约简算法具有一致性。(4)对查找“黑广播”信息系统进行属性约简。应用本文提出的约简算法,获得了与监测经验相吻合,更实用、精简的判断条件。
[Abstract]:Genetic reduction algorithm is one of rough set reduction algorithms. In order to keep the classification ability of information system not a prerequisite, the redundant attributes are deleted, and the decision rules or classification rules with stronger generalization ability are obtained. Attribute reduction plays an important role in avoiding dimension disaster and improving the efficiency of the algorithm. Illegal broadcasting, commonly known as "black broadcasting", can affect radio order and even interfere with the normal use of aviation frequency bands. The broadcast that propagandizes false, vulgar advertisement is one of common "black broadcast". Intelligent monitoring is one of the effective measures to search for "black broadcast". It is an important feature of most "black broadcasts" that the voice is the main feature of the blackout without punctuation. The main contents of this paper are as follows: (1) an algorithm of full point timing recognition is proposed. After preprocessing and feature extraction, the hidden Markov models of the full and untimed FM broadcasting signals are established, and the effectiveness of the proposed method is verified by using the measured data. The information system of "black broadcast" is constructed based on the conventional monitoring index. (2) A genetic reduction algorithm for variable length chromosomes based on equivalence relation information system is proposed. Taking the variable length chromosome genetic algorithm as the basic frame, the fitness function is constructed based on the row and column importance of the simplified binary difference matrix. By introducing the adjustment function, the evolutionary process adaptively performs crossover and mutation operations, and the individual selectively performs genetic operations. The genetic reduction algorithm is suitable for compatible decision table and incompatible decision table. In this paper, attribute reduction for multiple UCI data sets is carried out, and the effectiveness and superiority of the proposed algorithm are verified by experimental comparison. (3) A genetic reduction algorithm for variable length chromosomes based on the information system of dominance equivalence relationship is proposed. In this paper, the fitness function is designed based on the positive domain of the information system of the advantage equivalence relation, and the attribute reduction of the decision table with the condition attribute value having partial order relation and the decision attribute value being the equivalent relation is realized. The experimental results show that the proposed algorithm is consistent with the traditional reduction algorithm. (4) attribute reduction is carried out for the search "black broadcast" information system. By using the reduction algorithm proposed in this paper, a more practical and concise judgment condition is obtained, which is consistent with the monitoring experience.
【学位授予单位】：西华大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18

【参考文献】