摘要
通过分析某银行网点19天所有客户到访的真实记录,挖掘银行客户到访的间隔时间分布、单日客户到访间隔时间分布、细分客户到访间隔时间分布等统计特征,发现银行客户的到访蕴含着胖尾的统计规律,并不是先前排队论假设的泊松过程。统计结果显示,人类访问银行的行为具有明显偏离泊松分布的胖尾特性,幂指数在2～3的范围内。这一结果为针对幂律事件间隔分布的排队理论的建立奠定了实证基础,为下一步银行排队问题的分析作了探索。
Through analyzing the real data of bank customers visiting in a bank outlet in 19 days, it was discovered that there were many forms of non-Poisson characters appearing in the data of bank customers visiting,such as inter-arrival time distribution of all customers in 19 days, total customers visiting in one day, customers visiting for personal banking business in 19 days and customers visiting for corporate banking business in 19 days. These characters were different from those in the hypothesis of queuing theory： the coming of customers could be well approximated by Poisson processes. These distributions denoted the pattern of bank customers visiting follows non-Poisson statistics or heavy tailed distribution. It is found that most of these distribution exponents are between 2 and 3. This result establishes the empical foundation of new queuing theory with power-law inter-arrival time distribution, and also explores the analysis of bank queuing in further studies.
出处
《上海理工大学学报》
CAS
北大核心
2012年第3期252-256,共5页
Journal of University of Shanghai For Science and Technology
基金
国家自然科学基金资助项目(70871082)
上海市重点学科建设资助项目(S30504)
上海市研究生创新基金资助项目(JWCXSL1022)
关键词
人类动力学
幂律分布
排队论
间隔时间分布
阵发性
human dynamics
power-law distribution
queuing theory
inter-arrival time distribution,
bursts
