基于随机森林算法的长沙市入境游客热点区域识别

Identification of area of interest for inbound tourists in Changsha based on random forest algorithm

当前游客热点识别方法主要依赖于游客的位置信息数据,未充分考虑地理因素对游客活动的影响,在地理意义上的解释力尚有不足。本文基于Flickr平台获取的带有地理标记的照片及其属性信息,结合多源时空数据,使用随机森林算法识别入境游客旅游热点区域(areaofinterest,AOI),探究长沙市入境游客AOI特征及成因;并与传统的密度聚类算法(P-DBSCAN)进行对比评价。结果表明:(1)随机森林算法在区域整体热度识别中表现优异,部分AOI边界具有明确的地理意义;(2)入境游客的整体分布格局较为稳定,核心区外围受距离影响明显,发展潜力不足,距景点400 m处为影响入境游客分布的临界点。本文为研究长沙市及类似城市的入境旅游热点区域和旅游资源配置提供了新视角。

Current methods for identifying areas of interest(AOIs)in tourism research often overlook less popular cities and primarily rely on tourist location data.These methods do not adequately account for the influence of geographical factors on tourist activities,resulting in limited geographical explanatory power.This study integrates the relationship between tourism activities and their influencing factors into the AOI identification process.It examines variations in tourist distribution density from a microscopic perspective,and establishes a reasonable threshold to clearly demarcate AOIs from non-AOIs.Utilizing machine learning techniques,this study establishes a data mapping relationship between tourist distribution and its influencing factors to perform AOI identification.The aim is to precisely identify hotspots of inbound tourist activities at a 30-meter grid scale.The study area is divided into basic grids,and the presence or absence of tourists in these grids is treated as two types of samples.A decision tree is constructed based on the relationship between grid characteristics and these samples,with prediction outcomes determined through the results of the decision tree.The question of tourist presence is thus transformed into a probability issue,using the likelihood of tourist presence to represent regional differences in tourist activity popularity.Data from geotagged photos and their attributes from the Flickr platform are used in conjunction with spatiotemporal data that quantify levels of tourist facilities,services,and resources.The random forest(RF)algorithm is then to identify inbound AOIs.Results are compared with those from a density-based spatial clustering of applications with noise for geotagged photos(P-DBSCAN)to investigate the characteristics and causes of inbound AOIs in Changsha.The findings indicate that the RF algorithm effectively identifies overall regional heat,offering richer information,broader coverage,and some AOI boundaries with clear geographical significance.In Changsha,AOIs are concentrated in three primary areas.The historical and cultural AOI serves as the core area visited by inbound tourists.Fewer AOIs are found outside this core area,mostly coinciding with local leisure and tourism zones,suitable for local recreation,travel,and shopping.The overall distribution pattern of inbound tourists remains relatively stable,with the periphery of the core area is significantly influenced by distance,indicating suboptimal tourism development potential.The critical point affecting inbound tourist distribution lies 400 meters outside attractions.Infrastructure and natural conditions exert minimal constraints on the distribution of inbound tourists.Urban tourism managers should focus on enhancing the attractiveness of popular sites and improving detailed tourist experiences.Applying the RF algorithm to inbound tourist studies compensates for the limitations of clustering algorithms,distinguishes regional popularity variations and their causes in detail,and thereby provides targeted insights.The model’s reliance is not rely solely on tourist data,avoiding certain issues related to data representativeness.The predictive outcomes can offer theoretical foundations and guidance for urban tourism planners,enriching the research content on urban inbound tourism.However,the study has some limitations;for instance,the spatial behaviors of inbound tourists result from a combination of multiple factors.This analysis only considered objective factors,lacking an in-depth exploration of subjective elements,which requires further investigation in future studies.