摘 要:随着消费者的消费能力逐渐增强以及旅游信息不透明程度逐年下降，消费者 的旅游意愿、消费行为逐渐变得难以预测，传统的旅游模式已经不能完全满足游 客的需求，而根据消费者的个人喜好、景点、天气、交通等维度定制合适的、受 欢迎的包车游路线并精准地向消费者推荐（即所谓的精品旅行服务）就既有广泛 的市场前景，又富有挑战性.本文针对精品旅行服务成单预测问题，提出了一种基 于多分类器和贝叶斯理论确定动态权重的客户成单预测集成模型.论文中首先使 用 K-means 算法和 ChiMerge 算法分别对类别型变量和数值型变量进行合并，然 后为了消除不平衡数据集的影响，借助 AUC 准则构造损失函数，优化了 Logistics 回归分类模型，为了进一步提高预测效果，又采用贝叶斯方法确定各个分类器的 动态权重，构建了精品旅行服务成单预测的贝叶斯集成模型.测试结果表明: 贝叶 斯集成模型的成单预测准确率高达 97.86%，结果优于单模型和其它集成模型.

该论文有图 6 幅，表 3 个，参考文献 13 篇. 

关键词：K-means 算法 ChiMerge 算法 Logistics 回归 贝叶斯方法

STUDY ON THE ORDER PREDICTION OF FINE QUALITY SERVICE FOR TRAVEL

ABSTRACT:As consumers’ consumption capacity gradually increases and the degree of opacity of travel information decreases year by year, consumers’ willingness to travel and consumer behavior become increasingly unpredictable. The traditional tourism model can no longer fully meet the needs of tourists, and is based on inpidual preferences of consumers. Dimensions, attractions, weather, transportation, etc. Customizing suitable, popular chartered tours and accurately recommending to consumers (so-called fine quality service for travel) have both broad market prospects and challenges. In this paper, aiming at the single-prediction problem of exquisite travel services, this paper proposes a customer-specific predictive integration model based on multiple classifiers and Bayesian theory to determine dynamic weights. In the thesis, K-means algorithm and ChiMerge algorithm are used to merge categorical variables and numerical variables respectively. Then, in order to eliminate the influence of unbalanced data sets, a logistic regression classification model is established based on the AUC criteria. In order to improve the forecasting effect, Bayesian method was used to determine the dynamic weights of each classifier, and an Bayesian integrated model on the order prediction of fine quality service for travel was constructed. The test results show that: the prediction accuracy of Bayesian integration model is up to 97.86%, and which is superior to the single model and other integrated models.

The paper has 6 figures, 3 tables and 13 references.

Key Words: K-means algorithm ChiMerge algorithm Logistics regression Bayes method

目 录

摘 要 I

ABSTRACT II

1 引言 1

1.1 研究的目的和意义 1

1.2 研究的内容和方法 1

2 模型准备 2

2.1 数据的来源 2

2.2