Border tourist flow prediction and attraction recommendation based on improved SVD with MPA

Abstract

As border tourism grows rapidly, the need for accurate tourist flow prediction and personalized attraction recommendation has become increasingly important to enhance the quality of tourism services. This study proposes an improved Singular Value Decomposition method optimized by the Marine Predators Algorithm. The Marine Predators Algorithm optimizes the dimensionality reduction parameters of Singular Value Decomposition, introduces a time decay factor and holiday adjustment function to improve the accuracy of tourist flow prediction, and combines a time-aware factor to achieve dynamic personalized recommendations. In practical scenario tests, the model achieves an Absolute Mean Error of 2.3% in tourist flow prediction on regular workdays, and maintains a prediction accuracy above 88% during holidays. At the same time, the model also achieves a user satisfaction rate of 88.9% in the recommendation module, and a coverage score of 85.6% for attractions. These results indicate that the model has significant advantages in tourist flow prediction and personalized recommendation for border tourism attractions, effectively improving prediction accuracy and recommendation quality. The study provides a new technical solution for intelligent management of border tourism, which is of great significance for improving tourism service quality and tourist satisfaction.