【作者】 段永桓；

【导师】 王洪礼；

【作者基本信息】 天津大学 ， 管理科学与工程， 2011， 博士

【摘要】 旅游业是海南省国际旅游岛的核心产业,为了进行更加科学的旅游规划设计(特别是对旅游业的投资决策),提高资产利用率,保证海南旅游业的健康蓬勃的发展,必须对海南旅游市场作出准确的预测:一是通过对各种影响因素的分析,进行预测(各种回归分析);另一种是通过自身的发展规律,进行预测(时间序列)。本文挑选影响旅游市场的关键因素,首次采用随机森林RFA法和Copula法,对海南旅游业进行预测;考虑到要把海南目前的以”观光旅游”为主,逐步转变为以”休闲旅游”为主,高尔夫旅游是重要因素,为此,本文还首次用非线性时间序列法,对高尔夫产业收入进行预测,具体如下:首先找出影响海南旅游的主要因素:全国城镇居民人均可支配收入、固定资产投资、海南省酒店数量、公路通车、铁路通车里程、民航航线里程以及高尔夫球洞数量,再计算它们的Kendall秩相关系数和互信息指数,判断它们和旅游收入的相关性,再采用随机森林RFA法拟合它们之间的函数关系,然后计算各因素的影响程度,结果表明居民人均可支配收入影响最大,酒店数量次之。然后,挑选最有影响力的因素,用Copulafa法拟合它们的概率分布,确定其相关结构为frank Copula,再结合RFA法的拟合函数,利用蒙特卡罗法预测出海南旅游收入的概率分布函数,计算出风险度量VaR和条件风险值。结果表明:2012年海南旅游收入超过295亿元的可能性为80%,而超过310亿元的可能性为20%。对高尔夫产业的预测,是以海南博鳌亚洲论坛国际会议中心高尔夫球会每月收入为例,先对其对数差分数据进行独立性检验和非线性检验,再利用自相关函数和偏相关函数以及AIC准则,选择差分+ARMA(3,2)模型,用极大似然法估计参数,最后预测了博鳌高尔夫球会月收入均值和95%置信空间更多还原

【Abstract】 Tourism is an important pillar industry of international island of Hainan. Mar-ket prediction is essential for tourism planning and design, especially for investmentdecision. It also do good to asset utilization ratio as well as healthy and vigorous de-velopment of junketing. There are two ways of prediction, one is finished by analysisof inffuencing factors, the other by finding its own laws. Key inffuencing factors wereselected. Based on Random Forest and Copula, Hainan travel industry was forecasted.Golf plays an essential role in transition from sightseeing to leisure tour. Nonlinear timeseries was applied to Golf estate prediction, as structured below:Firstly the national average disposable income of urban residents, fixed asset in-vestment, hotels numbers, road traffc mileage, railway traffc mileage, airplane traffcmileage and Golf holes in Hainan were picked out. Rank correlation coeffcients andmutual information indexes between Hainan tourism income and them were calculatedto order dependence. The function among them was fitted by Random Forest and vari-able importance was computed. The result showed that the national average disposableincome has the most inffuence on travel income and hotel number came second.Secondly the Copula of five key factors was fitted as Frank Copula. Combiningwith Random Forest fitted function between tour income and inffuencing factors, prob-ability distribution function of future income was simulated by Monte Carlo. VaR andconditional VaR were figured out. It appeared that tour gainings in Hainan in 2012 has80% probability to exceed 29.5 billion, while 20% probability to exceed 31 billion.Thirdly Golf club in BFA International Convention Center was taken as an examplefor Golf estate forecasting. Independence and nonlinear tests are finished on logarithmdifference transformation data. ARMA(3,2) was taken as last model by autocorrelationfunction, partial correlation function and AIC criterion. Parameters were estimated us-ing maximum likelihood estimation. The mean and 95% confidence interval of Monthlyrevenues in Boao golf club were predicted.更多还原