<![CDATA[This research aims to develop a predictive recommendation model for tourist destinations in Indonesia using a hybrid approach that integrates TF-IDF and Cosine Similarity-based content analysis with Neural Network modeling. The main problem addressed is the high complexity in determining relevant tourist destinations due to the numerous choices and diversity of characteristics among destinations. The dataset used consists of 935 tourist destination data points, including text and numeric attributes. The TF-IDF method is used to extract features from destination descriptions, followed by Cosine Similarity to calculate their compatibility with user preferences. Meanwhile, the Neural Network processes numeric features such as category, province, and description length to generate additional relevance prediction scores. These two scores are combined through a hybrid approach to achieve more accurate recommendation results. Evaluation results show that the hybrid model performs very well with Precision@250 of 0.90 and Recall@250 of 0.97, indicating that the model not only selects relevant destinations but also successfully captures nearly all destinations matching user preferences. Overall, this hybrid approach provides more comprehensive recommendations compared to single methods and has the potential to be implemented in a national-scale tourism recommendation system.]]>
Copyrights © 2026
