基于协同演化遗传算法的个体特征信息识别技术方法  

Individual Feature Information Recognition Methods Based on Collaborative Evolutionary Genetic Algorithm

作　　者：关金金[1] GUAN Jinjin(School of Information Engineering,Anhui Business and Technology College,Hefei Anhui 231131,China)

机构地区：[1]安徽工商职业学院信息工程学院,安徽合肥231131

出　　处：《长沙大学学报》2024年第5期31-38,共8页Journal of Changsha University

基　　金：安徽省高校自然科学研究项目“模式匹配算法在高校采购风险防控中的应用研究”(KJ2021ZD0174);安徽省职业与成人教育学会教育科研规划重点课题“OBE视域下BOPPPS教学模型在‘Hadoop大数据平台构建与应用’课程中的改革与训练研究”(AZCJ2023009)。

摘　　要：在当前的个体特征信息分析过程中,依靠单一的遗传算法进行特征识别,只考虑了个体对自然环境的适应情况,使得最终识别结果AUC值较低。因此,提出基于协同演化遗传算法的个体特征信息识别技术(CEGA-IFIR)。运用贝叶斯网络构造最优分类器,从大数据中挖掘出个体数据,再展开局部低秩矩阵补全(LRMC)集成学习,实现个体数据集中缺失数据的插补处理。以信息熵概念为核心,设计个体特征向量信息增益评估函数。以最大特征信息增益为目标,结合协同演化算法和遗传算法进行个体特征信息识别求解,充分考虑多个个体在进化过程中的相互影响、相互适应特点,输出有效的特征信息识别结果。实验结果表明:CEGA-IFIR方法应用后,所得个体特征信息识别结果的AUC值相较于两种对比方法的AUC值(分别为0.74和0.61)更高,最大值达到0.93,满足了预期设计要求。The current method of analyzing individual feature information merely relies on a single genetic algorithm for feature recognition,and only considers the individual’s adaptation to the natural environment which leads to a lower AUC value in the final recognition result.Therefore,a collaborative evolutionary genetic algorithm based individual feature information recognition technology(CEGA-IFIR)is proposed.The proposed method uses Bayesian networks to construct optimal classifiers,applies them to mine individual data from big data,and then expands local low rank matrix completion(LRMC)ensemble learning in order to achieve interpolation processing of missing data in individual datasets.Based on the core concept of information entropy,an individual feature vector information gain evaluation function is designed.To maximize feature information gain,combining with collaborative evolution algorithm and genetic algorithm,individual feature information recognition is solved,fully considering the mutual influence and adaptation characteristics of multiple individuals in the evolution process,and outputting effective feature information recognition results.The experimental results show that based on the CEGA-IFIR method,the AUC value of the individual feature information recognition results is higher than the 0.74 and 0.61 of the two comparison methods,with a maximum value of 0.93,meeting the expected design requirements.

关 键 词：协同演化 遗传算法 个体特征 缺失值插补 遗传编码 信息识别 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]