线上旅游产品销量的影响因素分析和预测

Analysis and Forecasting of Influencing Factors of Online Travel Product Sales

摘要: 随着信息技术的飞速发展，在线旅游成为一种新兴的商业模式，在线旅游网站具有非常丰富的产品信息，包括产品价格、消费者评论、旅行社介绍等，其中每一项指标都会对线上旅游产品的销量产生不同程度的影响。本论文使用八爪鱼采集器对携程旅行网的在线产品的相关信息进行收集。使用Logit定序回归的方法，筛选出影响历史销量的12个关键因素，包括目的地类型、旅游类型、产品评分、点评数量、产品价格、旅行社评分、精致小团、赠取消险、无购物、攻略完备、立即确认和限时促销。最后，分别使用决策树模型和随机森林模型对产品销量进行了预测，参数调优后随机森林模型的AUC为0.9749。相较于决策树模型(AUC = 0.9190)，随机森林模型显示出更高的预测精度，能够更好地区分正例和负例样本，具备更出色的分类性能。

Abstract: With the rapid development of information technology, online tourism has become an emerging business model. Online travel websites have a wealth of product information, including product prices, consumer reviews, travel agency introductions, and more. Each of these indicators will have varying degrees of impact on the sales of online travel products. This thesis utilizes the Octopus Collector to gather relevant information about the online products from Ctrip.com, a renowned travel website. By using Logit fixed-order regression, 12 key factors that affect his-torical sales were filtered out. These factors include destination type, tour type, product rating, number of reviews, product price, travel agent rating, exquisite small group, complimentary cancellation insurance, no shopping, complete strategy, immediate confirmation, and limited-time promotion. Finally, the decision tree model and random forest model were utilized to predict product sales. After parameter tuning, the random forest model achieved an AUC of 0.9749. When compared to the decision tree model (AUC = 0.9190), the random forest model demonstrates higher prediction accuracy, better differentiation between positive and negative samples, and superior classification performance.