Research On Recommendation Method Of Tourist Attractions Based On Weibo Big Data And Machine Learning Algorithms

Accurate recommendation of tourist attractions is conducive to improving efficiency and experience of tour.However,the choice of the factors and algorithms of scenic recommendation will affect the accuracy of scenic recommendation.Aiming at the problems of sparse data,inadequate tourism factors and low recommendation accuracy in existing tourism recommendation research,this paper proposes a scenic spot recommendation method based on micro-blog big data and machine learning algorithm,which realizes precise and personalized tourism scenic spot recommendation by utilizing the characteristics of personalized expression of micro-blog data,strong current situation and intelligent prediction function of machine learning.The main work and results of the thesis are as follows:(1)The microblog data has the characteristics of large amount of data,rich semantics,real user’s thoughts and easy access,which can alleviate the problem of data sparseness of traditional travel website data for travel recommendation.Therefore,this article first uses the Python crawler to obtain Sina Weibo related to the attraction,and classifies and cleans the acquired data for the recommendation research of tourist attractions.(2)Secondly,this paper explores a wealth of tourism characteristics.The typical travel recommendation algorithm excavates tourism feature factors from attractions,tourists,etc.,without considering the context information such as the travel time and travel season of visitors to the destination,and they can help to understand the user’s travel preferences from different angles.This paper extracts six characteristic factors of scenic spot location,scenic spot fare,scenic spot level,main category,sub-category and basic type from the perspective of tourist attractions.From the perspective of tourists,the method of statistical analysis is used to extract gender and age.Four characteristic factors of age,source and source;from the perspective of context-aware information,three feature factors of transit duration,season and month are extracted by means of geographic concentration index.Based on this,a rich library of tourism feature factors is established,and multiple characteristics are combined to provide support for reliable prediction.(3)Thirdly,for the data sparse and cold start problem of collaborative filtering recommendation algorithm,this paper introduces machine learning algorithm and combines the proposed multi-feature tourism factor to construct dynamic scenic spot prediction model named RFPAP(Random Forest Preferred Attraction Prediction)and NNPAP(Neural Networks Preferred Attractions Prediction).The experimental results show that the RFPAP and NNPAP methods can overcome the data sparsity problem,and obtain the accuracy of 89.61% and 89.51% respectively,and the RFPAP method is superior to the NNPAP method and has stronger generalization ability.(4)Fourthly,this paper uses the FP-Growth algorithm to construct the model named FP-Growth AA(FP-Growth Attraction Association).This model can be used to efficiently mine the association rules between the attractions in the microblog big data.The experimental results show that by mining the relationship between the attractions selected by tourists,it can provide effective information for tourism travel decisions.(5)Finally,an attraction recommendation method based on RFPAP and FPGrowth AA model is proposed.It can not only predict tourists’ preference for scenic spots,but also excavate scenic spots with strong relevance to preference spots.It can be recommended to target tourists according to confidence ranking,which effectively improves the accuracy of personalized recommendation of scenic spots and has strong generalization ability.