流数据作为一种新型数据,在各个领域均有应用,其快速、大量及持续不断的特点使得单遍精准扫描成为在线学习算法的必备特质.在流数据不断产生过程中,往往会发生概念漂移,目前对于概念漂移节点检测的研究相对成熟,然而实际问题中学习环境...流数据作为一种新型数据,在各个领域均有应用,其快速、大量及持续不断的特点使得单遍精准扫描成为在线学习算法的必备特质.在流数据不断产生过程中,往往会发生概念漂移,目前对于概念漂移节点检测的研究相对成熟,然而实际问题中学习环境因素朝不同方向发展往往会导致流数据中概念漂移类别的多样性,这给流数据挖掘及在线学习带来了新的挑战.针对这个问题,提出一种基于时序窗口的概念漂移类别检测(concept drift class detection based on time window,CD-TW)方法.该方法借助栈和队列对流数据进行存取,借助窗口机制对流数据进行分块学习.首先创建2个分别加载历史数据和当前数据的基础节点时序窗口,通过比较二者所包含数据的分布变化情况来检测概念漂移节点.然后创建加载漂移节点后部分数据的跨度时序窗口,通过分析该窗口中数据分布的稳定性检测漂移跨度,进而判断概念漂移类别.实验结果表明该方法不仅能够精确定位概念漂移节点,同时在漂移类别判断方面也表现出良好性能.展开更多
For a risk process R_u(t) = u + ct- X(t), t≥0, where u≥0 is the initial capital, c > 0 is the premium rate and X(t), t≥0 is an aggregate claim process, we investigate the probability of the Parisian ruin P_S(u, ...For a risk process R_u(t) = u + ct- X(t), t≥0, where u≥0 is the initial capital, c > 0 is the premium rate and X(t), t≥0 is an aggregate claim process, we investigate the probability of the Parisian ruin P_S(u, T_u) = P{inf (t∈[0,S]_(s∈[t,t+T_u])) sup R_u(s) < 0}, S, T_u > 0.For X being a general Gaussian process we derive approximations of P_S(u, T_u) as u →∞. As a by-product, we obtain the tail asymptotic behaviour of the infimum of a standard Brownian motion with drift over a finite-time interval.展开更多
文摘流数据作为一种新型数据,在各个领域均有应用,其快速、大量及持续不断的特点使得单遍精准扫描成为在线学习算法的必备特质.在流数据不断产生过程中,往往会发生概念漂移,目前对于概念漂移节点检测的研究相对成熟,然而实际问题中学习环境因素朝不同方向发展往往会导致流数据中概念漂移类别的多样性,这给流数据挖掘及在线学习带来了新的挑战.针对这个问题,提出一种基于时序窗口的概念漂移类别检测(concept drift class detection based on time window,CD-TW)方法.该方法借助栈和队列对流数据进行存取,借助窗口机制对流数据进行分块学习.首先创建2个分别加载历史数据和当前数据的基础节点时序窗口,通过比较二者所包含数据的分布变化情况来检测概念漂移节点.然后创建加载漂移节点后部分数据的跨度时序窗口,通过分析该窗口中数据分布的稳定性检测漂移跨度,进而判断概念漂移类别.实验结果表明该方法不仅能够精确定位概念漂移节点,同时在漂移类别判断方面也表现出良好性能.
基金the Swiss National Science Foundation (Grant No. 200021140633/1)the project Risk Analysis, Ruin and Extremes (an FP7 Marie Curie International Research Staff Exchange Scheme Fellowship) (Grant No. 318984)Narodowe Centrum Nauki (Grant No. 2013/09/B/ST1/01778 (2014-2016))
文摘For a risk process R_u(t) = u + ct- X(t), t≥0, where u≥0 is the initial capital, c > 0 is the premium rate and X(t), t≥0 is an aggregate claim process, we investigate the probability of the Parisian ruin P_S(u, T_u) = P{inf (t∈[0,S]_(s∈[t,t+T_u])) sup R_u(s) < 0}, S, T_u > 0.For X being a general Gaussian process we derive approximations of P_S(u, T_u) as u →∞. As a by-product, we obtain the tail asymptotic behaviour of the infimum of a standard Brownian motion with drift over a finite-time interval.