最小化误差平方和k-means初始聚类中心优化方法

传统的k-均值算法对初始聚类中心和孤立点敏感,文中以最大程度地减少误差平方和为基本思想,提出一种最大化减少当前误差平方和的k-means初始聚类中心优化方法。在初始聚类中心选择阶段,每次增加聚类中心时,计算所有数据点作为当前聚类中心能够减少的误差平方和,选择能够最大化减少误差平方和的数据点作为聚类初始中心。利用真实数据集,同其他算法进行对比,实验结果表明该方法在选择初始聚类中心方面能够有效地减少聚类的迭代次数,提高聚类质量。同时人工模拟数据表明该方法对孤立点相对不敏感。     

Email Sentiment Analysis Through k-Means Labeling and Support Vector Machine Classification

Sentiment analysis for social media and online document has been a burgeoning area in text mining for the last decade. However, Email sentiment analysis has not been studied and examined thoroughly even though it is one of the most ubiquitous means of communication. In this research, a hybrid sentiment analysis framework for Email data using term frequency-inverse document frequency term weighting model for feature extraction, and k-means labeling combined with support vector machine classifier for sentiment classification is proposed. Empirical results indicate comparatively better classification results with the proposed framework than other combinations.     

基于聚类与分类结合的多示例预测算法

针对训练包不含标签的无监督多示例问题,本文提出了聚类和分类结合的多示例预测算法。首先利用多示例聚类算法完成无监督多示例学习的聚类任务,并根据聚类结果,将各个簇中的每个包转换成相应的k维特征向量。在标准多示例预测模型和一般性多示例预测模型上进行实验,可以得到较高的预测准确度,与其它多示例预测算法相比,本文算法具有较好的性能。     

Improved rough k-means clustering algorithm based on weighted distance measure with Gaussian function

Rough k-means clustering algorithm and its extensions are introduced and successfully applied to real-life data where clusters do not necessarily have crisp boundaries. Experiments with the rough k-means clustering algorithm have shown that it provides a reasonable set of lower and upper bounds for a given dataset. However, the same weight was used for all the data objects in a lower or upper approximate set when computing the new centre for each cluster while the different impacts of the objects in a same approximation were ignored. An improved rough k-means clustering based on weighted distance measure with Gaussian function is proposed in this paper. The validity of this algorithm is demonstrated by simulation and experimental analysis.     

基于支持向量机和互联网信息修正的空间负荷预测方法

为提高电网规划阶段的空间负荷预测精度,提出了一种基于支持向量机和互联网信息修正的空间负荷预测（spatial load forecasting,SLF）方法,该方法分为3个步骤：一是基于k-均值聚类分析和支持向量回归模型得到地块负荷初始预测值;二是基于地块负荷历史数据计算负荷实际值与初始预测值之间的偏差;三是针对这些偏差,利用搜索引擎获取互联网信息,识别造成偏差的不确定事件,包括元胞中新增大负荷事件和元胞中企业营收增长率突变事件。定性分析事件对空间负荷的影响,并建立这两类事件与其造成的影响之间的分类事件影响定量模型,基于该模型对地块负荷初始预测值进行修正,得到规划区域内的地块负荷预测值。通过对北京某地区进行算例验证,结果表明该方法可以提高预测精度,可用于配电网以及能源互联网规划中的空间负荷预测。     

二维卷积非负矩阵分解的初值确定混合算法

为解决二维卷积非负矩阵分解算法存在初值敏感,且传统随机初始化确定的初始值容易使算法收敛到结果相对不好的局部最优值的问题,通过结合K均值聚类、奇异值分解和主成分分析方法,提出了一种适用于二维卷积非负矩阵分解初始值确定的混合算法.首先,利用K均值聚类方法得到聚类中心作为系数矩阵(H矩阵)的初始值,避开了传统初始化不确定系数矩阵带来的分解结果不唯一问题;其次,考虑到相比一维卷积非负矩阵分解算法,二维卷积非负矩阵分解算法的基矩阵(W矩阵)个数更多,利用奇异值分解和主成分分析方法交替产生基矩阵的初始值,克服了单个算法产生的初始化误差问题.在相同参数环境下将本文算法和现有初始化算法的分解收敛性能进行对比实验,结果表明本文算法相比其他同类算法具有更好的分解性能并具有更好的收敛性.进一步加入噪声进行实验,在白噪声为-1 dB～10 dB的不同信噪比环境下,本文算法均能快速实现信号的分离,对于噪声数据具有很强的鲁棒性.采用混合算法确定初值,更有利于实现二维卷积非负矩阵分解的实时性和高性能.     

液晶屏线路中导电粒子压合的自动光学检测研究

在薄膜晶体管液晶显示器线路检测中,常通过对线路中的导电薄膜粒子的计数和定位实现其导电性的自动检测。为了解决窄边框线路中粒子密度增大带来的粒子重叠问题,提出一种采用微分干涉成像和掩模法结合k均值聚类的算法,在分离出粒子的亮、暗部后,结合图像熵值和粒子的凸性准确分割出粒子。讨论了聚类簇选值的影响,通过不同粒子密度、不同粒子尺寸的样本检验本文算法,并与以往的梯度结合灰度的方法进行对比。结果表明:本文算法在粒子密度较小的区域能达到92.6%的识别率,在粒子密度较大的区域也能达到86%的识别率,分别比梯度加灰度的方法提高了9.9%和42.7%。解决了粒子重叠的问题,并且对光场和成像效果有更好的鲁棒性。     

Multi-objective scheduling problems: Determination of pruned Pareto sets

There are often multiple competing objectives for industrial scheduling and production planning problems. Two practical methods are presented to efficiently identify promising solutions from among a Pareto optimal set for multi-objective scheduling problems. Generally, multi-objective optimization problems can be solved by combining the objectives into a single objective using equivalent cost conversions, utility theory, etc., or by determination of a Pareto optimal set. Pareto optimal sets or representative subsets can be found by using a multi-objective genetic algorithm or by other means. Then, in practice, the decision maker ultimately has to select one solution from this set for system implementation. However, the Pareto optimal set is often large and cumbersome, making the post-Pareto analysis phase potentially difficult, especially as the number of objectives increase. Our research involves the post Pareto analysis phase, and two methods are presented to filter the Pareto optimal set to determine a subset of promising or desirable solutions. The first method is pruning using non-numerical objective function ranking preferences. The second approach involves pruning by using data clustering. The k-means algorithm is used to find clusters of similar solutions in the Pareto optimal set. The clustered data allows the decision maker to have just k general solutions from which to choose. These methods are general, and they are demonstrated using two multi-objective problems involving the scheduling of the bottleneck operation of a printed wiring board manufacturing line and a more general scheduling problem.     

抽样改进加权核大数据谱聚类算法

经典谱聚类算法将数据聚类转为图划分问题,在分析其Normalized Cut函数与传统加权核k-means等价基础上,设计了一种基于抽样改进加权核k-means算法的大规模数据集谱聚类算法,算法通过加权核k-means迭代优化避免Laplacian矩阵特征分解的大量资源占用,通过随机映射得到近似奇异值分解,并由近似奇异向量确定各点数据权重及抽样概率,以此得到快速合理抽样,通过数据抽样并将聚类中心约束到抽样点生成的子空间中,避免全部核矩阵的使用,从而降低经典算法的时间空间复杂度。实验结果表明,改进算法在保持与经典算法相近精度基础上,大幅提高了聚类效率,实验验证了改进算法的有效性。     

基于粗糙集的混合属性数据聚类算法

传统聚类方法将对象严格地划分到某一类,但是很多时候边界对象不能被严格地划分。基于粗糙集的k-means聚类算法和基于粗糙集的leader聚类算法,利用粗糙集理论将数据对象划分到一个簇的上近似集或下近似集当中,提供了一种新的处理不确定性的视角,很好地解决了这种边界不确定问题。但其缺点是不能处理混合属性数据,聚类结果对初值有明显的依赖性。针对这些算法存在的不足,给出了一种适用于混合属性数据的距离定义,对初始值的选取提出了改进办法,提出了一种基于粗糙集的混合属性数据聚类算法。仿真实验证明,在不确定聚类簇数的情况下,该算法的聚类准确率比传统k-means算法明显提高。