改进K近邻和支持向量机相融合的天气识别  

Weather identification based on improved K nearest neighbor and support vector machine

作　　者：张红艳[1] 李茵茵[1] 万伟[2] 

机构地区：[1]广州中心气象台,广州510080 [2]广东广播电视大学计算机技术系,广州510091

出　　处：《计算机工程与应用》2014年第14期148-151,167,共5页Computer Engineering and Applications

基　　金：广东省气象局气象科技项目(No.2011B03;No.201007)

摘　　要：天气受到多种因素综合影响,具有时变性和不确定性,单一模型难以获得较高的识别正确率,为此,提出一种改进K近邻和支持向量机相融合的天气识别模型(IKNN-SVM)。首先计算待识别样本与超平面间距离,然后将距离与预设阈值进行比较,如果大于阈值,则采用支持向量机对天气进行识别,否则利用K近邻算法对天气进行识别,并引入样本密度对K近邻算法进行改进,最后采用仿真实验对模型性能进行测试。仿真结果表明,相对于单一的KNN或SVM,IKNN-SVM提高了天气识别正确率,较好地克服单一模型存在的缺陷。The weather which is affected by many factors is changeable and uncertain, single model is difficult to obtain high identification rate, therefore, this paper proposes a weather identification model（IKNN-SVM）based on improved K nearest neighbor and support vector machine. Firstly, the distance between of the testing sample and a hyper plane is cal-culated, then the distance is compared with the threshold, if distance is greater than the threshold, then support vector machine is used to identify the weather, otherwise the K nearest neighbor algorithm is used to identify the weather, and the sample density is introduced to solve the defects of K nearest neighbor algorithm, finally the simulation experiment is car-ried out to test on the performance of model. The simulation results show that, compared with the single KNN or SVM, IKNN-SVM has improved weather identification correct rate and can overcome the defects of the single model.

关 键 词：天气识别 支持向量机 K近邻 识别正确率 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]