Research On Tourist Flow Prediction And Personalized Recommendation Algorithms For Tourist Attractions

Accurate predictions of tourist flow at attractions can assist in the organization management and resource allocation of these attractions,while recommendations for tourist attractions can reduce the information asymmetry between tourists and attractions.Machine learning algorithms have received considerable attention for their excellent performance in nonlinear complex modeling,and they have been applied to time series prediction and personalized recommendation.In this paper,we conducted an in-depth study on tourist flow prediction and recommendation algorithms for tourist attractions and proposed a tourist flow prediction method based on kernel principal component analysis,binary multi-strategy marine predator algorithm,and regularized extreme learning machine,as well as a tourist attraction recommendation algorithm based on the multi-strategy marine predator algorithm optimized SVD++.The main research contents of this paper are as follows:First,a multi-strategy marine predator algorithm is proposed by integrating adaptive adjustment,group learning,and stagnation perturbation strategies.This algorithm efficiently utilizes the exploration and exploitation phases,effectively increases the diversity of search agents,and enhances global search capabilities.Simulation experiments on the CEC 2015 benchmark test functions show that the multi-strategy marine predator algorithm performs well in different variable dimensions compared to original algorithms,single-strategy improved algorithms,and similar algorithms,and passes the Friedman rank-sum test.Subsequently,a corresponding binary version of the algorithm is designed and applied to feature selection.Second,a tourist flow prediction model based on kernel principal component analysis,multi-strategy marine predator algorithm,and regularized extreme learning machine is constructed using tourist flow data.This method uses kernel principal component analysis for feature extraction of the web search index,binary multi-strategy marine predator algorithm for lag feature selection,and multi-strategy marine predator algorithm for optimizing the initialization of weights and biases in the regularized extreme learning machine.This paper conducts an empirical analysis of the Changbai Mountain 5A-level tourist attraction,incorporating multi-source data such as historical tourist flow,web search index,weather,holidays,and seasons,and verifies the predictive performance of the proposed method from the perspectives of information inclusion and model optimization.Third,a tourist attraction recommendation model based on the multi-strategy marine predator algorithm optimized SVD++ is constructed using tourist user attraction rating data,determining the optimal learning rate and regularization degree initialization parameters for the SVD++ algorithm.This paper filters tourism domain data from the Yelp dataset,constructs a tourist user attraction rating dataset,and verifies that the proposed method has certain advantages over collaborative filtering and other matrix factorization algorithms and that the multi-strategy marine predator algorithm can further optimize the recommendation performance of the SVD++ algorithm.In summary,this paper conducts research on the prediction of tourist flow at tourist attractions and the recommendation of tourist attractions,proposing a tourist flow prediction method based on kernel principal component analysis,binary multi-strategy marine predator algorithm,and regularized extreme learning machine.This method can accurately predict tourist flow at attractions,helping to efficiently allocate tourism resources,ensure the balance of supply and demand for tourism products and services,and can be extended to air quality and exchange rate prediction.Meanwhile,the multi-strategy marine predator optimized SVD++ tourist attraction recommendation algorithm proposed in this paper can meet the needs of tourists for attraction recommendations and enhance their travel experience.