Digital Tourism Recommendation and Route Planning Model Design Based on RippleNet and Improved GA

Abstract

Tourism recommendation and route planning are important applications of smart tourism. To achieve digital tourism that integrates tourist attraction recommendation and route planning, this study first integrates RippleNet, an item representation enhancement module, and a knowledge graph to construct a new tourist attraction recommendation model. Secondly, to address the deficiencies of the genetic algorithm, it improves the population initialization and searching ability and designs the route planning model between tourist attractions. The experimental results indicated that the research-designed recommendation model had superior results in the evaluation of average accuracy mean and receiver operating characteristic curve, the average accuracy mean was higher than 0.9, and the curve area reached 0.924. At the same time, the model's root-mean-square error was as low as 0.316, the average absolute error was as low as 0.247, and the maximum of the R-squared index reached 0.844, and the Huber loss The lowest value was 0.215. The combined metrics verified the superiority of the model. The mean inverse ranking and comprehensive coverage verified the recommended utility of the model. In addition, the improved genetic algorithm's hypervolume index and anti-generation distance evaluation index indicated the rationality of the improved strategy, with clear path planning results and high scores of path fluency and rationality. This research can enhance the level of tourism service function, realize the personalized customized service of tourism travel, and enrich the tourists' experience to promote the development of digital tourism economy.