广西各地级市城市职能分类探讨

城市职能是城市发展的重要标志之一.基于人工神经网络(ANN)中自组织特征映射网络(SOM)的聚类功能,提取广西各地级市城市行业从业人员指标,应用MATLAB6.5软件,对广西各地级市城市职能进行分类;同时对SOM人工神经网络分类方法的优缺点及城市职能的分类结果进行分析.     

超高频数据的日内效应调整方法研究

日内效应在金融高频数据研究中已被广泛证实,是一种日内周期性运动的动态效应,它影响了以微观金融指标为参数的计量模型的准确估计。基于金融超高频持续期数据,本文首先论述了日内效应调整的重要性,然后引入自适应映射(SOM)的方法对日内效应进行调整。SOM是一种基于神经网络学习的特征提取方法,能够动态识别高维数据中的结构特征,克服了静态调整方法的不足。最后通过建立基于自回归条件持续期模型(ACD)的蒙特卡罗模拟实验,比较了三种日内效应调整方法的效果。模拟结果表明SOM方法在日内效应调整中更为有效和稳定,特别适合大数据条件下的周期性结构分析。     

Growth hormone,somatomedins and men's health

According to current basic science as well as the clinical literature, human growth hormone has become an important topic in the field of anti-aging medicine. It is well known that the administration of human growth hormone compensates for the dwarfism syndrome in growth hormone-deficient children and growth hormone is also established as a substitution therapy for adults with pituitary deficiency. The effects mediated by growth hormone comprise an increase in muscle mass, a decrease in body fat, improved physical condition, oxygen consumption and overall quality of life, an improvement of the ratio of high-density lipoprotein to low-density lipoprotein, as well as an increase in bone density. Due to the insulin antagonistic effect of growth hormone as well as its insulin-like growth factor-1-mediated mitogenic effect, there is reason to doubt the safety of administering growth hormone outside the framework of approved indications. In an ongoing study, we are evaluating whether growth hormone and somatomedins might be of importance in the regulation of male sexual performance, including penile erection.     

基于蚁群算法的成本敏感线性集成多分类器的客户流失研究

针对电信客户流失预测问题的复杂性,融合自组织神经网络良好的连续属性值离散化优势、粗糙集理论出色的属性约简功能和蚁群优化算法全局的随机搜索特点,在模型集成技术和成本敏感学习理论的基础上,提出了一种新的基于蚁群算法的成本敏感线性集成多分类器的电信客户流失预测模型。构建该集成模型可分为4个阶段:(1)连续属性值的离散处理:利用自组织神经网络对连续属性值进行非监督离散化处理;(2)原始属性集的约简处理:使用粗糙集理论按属性重要性原则对离散属性进行约简;(3)子分类器的建立:分别使用NaiveBayes、Logistic回归、多层感知器和决策树等4种差异性很大的分类技术在约简属性集上建立4个对应的客户流失预测子分类器;(4)子分类器的集成:基于成本敏感学习理论,构建了4种不同的线性集成模型,采用蚁群优化算法求解集成模型的最优线性组合权重系数。将该模型应用于某电信客户流失预测,其实验结果表明该集成方法是可行且有效的。