Tourist Behavior Mining And Tourism Activity Recognition Based On Swarm-intelligent Sensing

With the development of the Internet and the popularization of intelligent terminal,Internet gradually integrated into the tourism industry,changing the traditional tourism business model and tourists' travel behavior.Tourist behaviors and tourism activities can not only reflect the tourists' travel characteristics directly,but also reflect features of tourist attractions.With the background of big data,using swarm-intelligent sensing technology to mine tourist behaviors and tourism activities,discovering potential demand of tourists has become a new application direction of tourism big data technology.The main work is completed as follows:(1)In mining of tourist behavior characteristics,the thesis firstly uses algorithm which is based on HITS(the Hyperlink-Induced Topic Search)model to recognize interesting attractions,secondly proposes a combined feature selection method(CI)of the chi-square test and information gain to classify tourist theme.Compared with the single feature selection method average accurate increased by 3.8%.Finally HCPFS is proposed to mine tourist behavior patterns which uses hierarchical clusteringto cluster tourists based on similarity of path trajectory and sojourn time,and mine tourist behavior patterns respectively.The validity of the behavior patterns mining is improved.(2)HP algorithm is proposed to reduce the impact of words that are not in HowNet or have low frequency in corpus which improves the traditional semantic similarity method based on HowNet and combines it with the method based on PMI.Compared with the single sentiment analysis method,accuracy of positive and negative emotion increased 6%and 5.3%on average respectively.RTBF_LA algorithm is proposed to predict the probability of secondary travel which overcoming the fitting.Compared with the Logistic,prediction accuracy increased nearly 8%.(3)Hot event mining algorithm LVCS is put forward,which combines text semantic similarity and feature words similarity,and uses spectral clustering to cluster texts.The problem that ambiguous sentences cannot be identified effectively by method based on the similarity of feature words is solved.Compared with the traditional method the cost price is reduced by an average of 28.1%.Tourism activity recognition algorithm(TAR KA)is proposed,by which microblogs which represent hot events nearby hot spots are classified to identify the tourist hot activity.Compared with KNN algorithm the average accuracy and F1 of TAR_KA is improved by 13.5%and 9.2%respectively with the feature selection methods CI.And the recognition accuracy of effective feature words is improved.(4)The thesis Designed and developed tourist behavior mining and tourism activity recognition system based on swarm-intelligent sensing.It is consist of tourist behavior mining module,secondary travel predict module and tourism activity recognition module.The high accuracy and good fault tolerance of the system has been verified.The system can meet the requirements of tourist behavior mining and tourism activity recognition basically.