基于随机森林和copula的海南旅游市场影响因素分析及预测 精灵论文.doc

基于随机森林和 copula 的海南旅游市场 影响因素分析及预测 段永桓1，王洪礼1，李胜朋2 5 （1. 天津大学管理学院，天津 300072； 2. 天津外国语大学基础课部，天津 300204） 摘要：旅游业是海南省的重要支柱产业，对旅游市场准确预测有助于合理规划海南国际旅游 岛建设。影响旅游市场因素很多，利用秩相关系数、互信息指数以及随机森林算法计算影响 因素对海南旅游收入的影响力大小。挑选重要影响力因素，分步拟合它们边缘分布和相关结 10 构 copula，根据随机森林拟合函数，以蒙特卡洛方法预测未来海南旅游收入概率分布，并 计算三种风险度量。结果表明，2012 年海南旅游收入超过 260.149 亿元的可能性为 80%， 而 超过 274.722 亿元的可能性才 20%。虽然突破 274.722 亿元可能性较小，但是如果有其它充 分的理由认为 2012 年海南旅游收入会超过 274.722 亿元，则它的平均收入是 1375.493 亿元。 关键词：旅游市场；随机森林；copula; 蒙特卡洛 15 中图分类号：X730 Influencing Factors Analysis and Prediction for Hainan Tourism Market by Random Forest and Copula DUAN Yongheng1, WANG Hongli1, LI Shengpeng2 20 (1. School of Management, Tianjin University, TianJin 300072; 2. Department of Fundamental Courses, Tianjin Foreign Studies University, TianJin 300204) Abstract: Tourism is an important pillar industry in Hainan province. Accurate prediction of tourism market is beneficial to construction of international island of Hainan. The market is influenced by many factors. The order of influence was calculated by rank correlation coefficient, 25 mutual information coefficient and random forest. Critical factors were pointed out and their margin distribution and copula were fitted. Based on fitted function from random forest, the probability distribution of future hainan traveling income was forecasted by Monte Carlo. Three risk measures were achieved. The result showed that for revenue in 2012 the chance of exceeding 26.0149 billion was eighty percent, while exceeding27.4722 billion was twenty percent. Although 30 exceeding 27.4722 billion has little possibility, the average revenue would be 137.5493 billion conditioning on it truly surpassed 27.4722 billion Keywords:tourism market; random forest; copula; Monte Carlo 0 引言 35 海南省是中国第一个提出建立国际旅游岛的省份，没有其它兄弟城市经验可供借鉴。世 界其它著名国际旅游岛如美国夏威夷群岛、澳大利亚大堡礁、西班牙巴利阿里群岛、马儿他 岛、墨西哥坎昆、巴哈马群岛等，由于地理、人文、政治等因素不同，各有各的发展模式， 没有统一的经验可供参考。为了合理规划海南国际旅游岛，科学预测旅游市场需求不可避免 [1]。 40 旅游市场及其变化状况直接影响到旅游目的地旅游经济的发展方向和发展规模，然而旅 游的强季节性及其产品的不可储存性使得旅游供给与需求间的矛盾较为突出，为避免过