Answers are grounded in this paper and cite their sources.
Hybrid model leverages Twitter and Flickr data to improve tourism trend prediction accuracy, suggesting significant enhancements in predictive capabilities.
This study addresses the limitations of traditional tourism trend prediction methods in effectively handling multimodal data by proposing a hybrid model based on social media multimodal data fusion — MM-HAL. This model combines the advantages of autoregressive integrated moving average (ARIMA) for capturing linear trends in time series with long short-term memory (LSTM) for modeling nonlinear features. The MM-HAL model leverages multi-modal data from Twitter and Flickr, including time series, text, images and geographic information, to predict tourism trends. The model also incorporates sentiment analysis of social media content to further improve prediction accuracy. Experimental results demonstrate that MM-HAL outperforms other models, such as LSTM, CNN-LSTM and Seq2Seq, with a mean absolute error (MAE) of 16.85 on the Twitter dataset and 16.50 on the Flickr dataset. These results show a significant improvement in prediction accuracy, achieving a 20% reduction in MAE compared to the LSTM model. Additionally, MM-HAL achieves a superior [Formula: see text] value of 0.89 on the Flickr dataset, indicating a better ability to explain data variance. The model also demonstrates relatively lower computational complexity, making it more efficient than Transformer models, particularly in large-scale data applications. By effectively combining ARIMA and LSTM for time series forecasting and sentiment analysis, MM-HAL offers an accurate, efficient and scalable solution for predicting tourism trends. Its flexibility and robustness make it suitable for applications in tourism management, marketing decision-making, and personalized recommendation systems.
No takes yet. Share an insight, caveat, or question.
Cheng et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: