Research On Inferring Tourists’ Behavioural Motives Based On Semantic Association Of Multi-Source Data

With the growth of tourism and the increasing competition in the tourism market,understanding tourists’ needs and motivations in order to provide more accurate tourism services has become the focus of tourism research.To understand tourist needs and behavioural characteristics,behavioural motivation mining and personal preference analysis using multiple sources of tourism data are required and used as a basis for decision making.The historical visitation behaviour of tourists is embedded in their personal preferences and characteristics,but the core issue of analysing tourists and their motivations from the perspective of geographical features is less researched,especially the use of the relationship between tourists and the geographical environment to tap into historical behaviour and thus dynamically capture tourists’ motivations.This thesis constructs a method for inferring visitor behavior motivation based on semantic association of multiple sources of data to address the problem that it is difficult to infer visitor behavior motivation from a large number of travelogue review texts and POI data and other data related to visitors’ historical behavior.The main studies are as follows:(1)A specific method of extracting tourists’ behavioural motives is studied,describing tourists’ behavioural motives from two perspectives: theme extraction and sentiment analysis,and digging deeper into the rich information on tourists’ historical behaviour contained in the travelogue review texts.Firstly,the CBOW model was used to optimise the LDA model,and the motivational themes were extracted from the review text according to the CBOW-LDA model,and the optimal number of themes was determined using the confusion degree;then the BERT model and Bi GRU bidirectional neural network were combined for sentiment analysis,and the attention mechanism was incorporated to weight the feature vectors to construct the Att-BERT-Bi GRU sentiment analysis model to analyse the sentiment bias of the travelogue review text and improve the accuracy of sentiment analysis.(2)A TS-FP-Growth algorithm incorporating geotemporal features is proposed,which constrains the FP-Growth algorithm from the perspective of geotemporal features of multisource tourism data.Firstly,the temporal and spatial features of multi-source tourism data are extracted,a geographical matrix is constructed according to the spatial features and a geographical threshold is set to filter the before and after items of association rules,then a time interval window is set according to the temporal features and the mining process is further constrained through the time window to obtain strong association rules related to tourists’ behavioural motives,which solves the problems of data redundancy and insufficient accuracy of results in the traditional FP-Growth algorithm for association mining.(3)A method of visitor motivation inference is investigated,combining behavioural motivation extraction and association analysis,which can effectively infer visitor behavioural motivation.The rich information left by users when they visit social networks is obtained,a sample dataset is constructed based on these multiple sources of tourism data,and the dataset is fed into the association rule algorithm for mining,and the association rules mined are interpreted and analysed to achieve inference of tourists’ behavioural motives.Comparative tests were designed to validate the proposed model and method.The experimental results indicate that the CBOW-LDA model proposed in this thesis,the improved Att-BERT-Bi GRU model and the TS-FP-Growth algorithm incorporating geographical features are all significantly optimised in terms of results,by setting parameters and influencing factors such as the environment,and comparing them with a variety of related models.In summary,the method proposed in this thesis has been significantly improved compared with other traditional methods,both in terms of accuracy and effectiveness of mining,confirming the effectiveness of the method of inferring tourists’ behavioural motives based on semantic association of multi-source data.