TRIP-Sense: A Data-Driven Travel Planning and Recommendation System

Abstract

Planning a trip in India is an intricate and time-consuming process involving the simultaneous evaluation of multiple factors such as destination, budget, trip duration, personal interests, and seasonal conditions. Existing travel platforms provide keyword-based filtering but fail to comprehend natural language or generate personalised, contextually aware itineraries. This paper presents TRIP-Sense, a hybrid travel recommendation system that integrates content-based filtering, collaborative filtering, and a lightweight natural language parser to deliver personalised destination suggestions, day-by-day AI-generated itineraries, budget analysis, and real-time interactive map visualisation. A multi-factor scoring engine is applied across 16 Indian destinations. The system optionally integrates the Claude API (Anthropic) for advanced generative itinerary responses. Evaluation demonstrates that the hybrid model outperforms individual filtering approaches, achieving over 90% NLP parsing accuracy and 84% overall recommendation accuracy. The complete system is delivered as a single-page web application using HTML, CSS, JavaScript, Leaflet.js, and the Overpass API with no server-side infrastructure required.

Introduction

Traditional travel platforms usually provide generic, popularity-based suggestions that do not match individual needs. To solve this, TRIP-Sense uses a hybrid recommendation approach combining content-based filtering and collaborative filtering to generate personalized destination recommendations. Users can input free-form queries, budget, interests, and trip duration, which the system processes using a custom NLP parser without external libraries.

The system is built as a fully client-side single-page web application, integrating multiple components: a query parser, recommendation engine, itinerary generator, map visualization (Leaflet + Overpass API), and a budget planner. It also includes a manually curated dataset of Indian tourist destinations with weather and travel information.

The hybrid recommendation model calculates scores using weighted contributions from content similarity and collaborative filtering (SVD), allowing more accurate and balanced suggestions. It returns top-ranked destinations along with interactive itineraries and cost breakdowns.

Performance results show strong effectiveness, with around 84% overall recommendation accuracy, high NLP extraction accuracy (up to 96.7% for interest detection), and strong top-3 recommendation reliability.

Conclusion

This paper has presented TRIP-Sense, a hybrid travel recommendation system that combines content-based filtering, collaborative filtering, and natural language processing within a fully client-side single-page web application. The system simplifies travel planning by accepting free-form natural language queries and returning personalised destination recommendations, day-by-day itineraries, budget breakdowns, and live interactive maps — all without server-side infrastructure. The NLP parser achieves over 90% field extraction accuracy. The recommendation engine correctly identifies the expected top destination in 80% of test cases and achieves 100% top-3 recall. AI-generated itineraries for Hyderabad, Goa, and Jaipur were rated above 4.3/5.0 by human evaluators on factual accuracy, relevance, and practical utility. Compared to server-dependent systems, TRIP-Sense provides a strong balance between performance, simplicity, and deployability. Future enhancements will include integration with real-time booking APIs (Google Maps, MakeMyTrip) for live pricing and availability; replacement of the rule-based NLP parser with a fine-tuned BERT intent extraction model supporting Hindi, Telugu, and Tamil queries; expansion of the knowledge base to 100+ Indian cities across all 29 states; development of a Progressive Web App (PWA) for offline mobile access; and incorporation of SHAP-based explainability tools to help users understand why each destination was recommended. Integration with EHR-style user travel profiles for continuous personalisation and deep learning approaches such as attention-based neural collaborative filtering are also planned for future work.

References

[1] C. C. Aggarwal, Recommender Systems: The Textbook. New York, NY, USA: Springer, 2016. [2] F. Ricci, L. Rokach, and B. Shapira, Eds., Recommender Systems Handbook, 2nd ed. New York, NY, USA: Springer, 2015. [3] R. Burke, “Hybrid recommender systems: Survey and experiments,” User Modeling and User-Adapted Interaction, vol. 12, no. 4, pp. 331–370, 2002. [4] G. Adomavicius and A. Tuzhilin, “Toward the next generation of recommender systems,” IEEE Trans. Knowledge and Data Engineering, vol. 17, no. 6, pp. 734–749, 2005. [5] D. Gavalas, C. Konstantopoulos, K. Mastakas, and G. Pantziou, “Mobile recommender systems in tourism,” Journal of Network and Computer Applications, vol. 39, pp. 319–333, 2014. [6] V. Aggarwal and R. Sharma, “A survey of travel recommendation systems,” International Journal of Computer Applications, vol. 162, no. 7, pp. 28–34, 2017. [7] J. L. Herlocker, J. A. Konstan, A. Borchers, and J. Riedl, “An algorithmic framework for performing collaborative filtering,” in Proc. ACM SIGIR, 1999, pp. 230–237. [8] X. Wang et al., “Benchmarking large language models as AI assistants,” arXiv preprint arXiv:2403.09014, 2024. [9] Scikit-learn, “Scikit-learn documentation.” [Online]. Available: https://scikit-learn.org/. [Accessed: Apr. 2025]. [10] Anthropic, “Claude API documentation.” [Online]. Available: https://docs.anthropic.com. [Accessed: Apr. 2025]. [11] Leaflet.js Contributors, “Leaflet — open-source JavaScript library for interactive maps.” [Online]. Available: https://leafletjs.com. [Accessed: Apr. 2025]. [12] Overpass API Contributors, “Overpass API documentation.” [Online]. Available: https://overpass-api.de. [Accessed: Apr. 2025]. [13] S. Choi, H. Jung, and M. Kim, \"Improving travel destination recommendations using cosine similarity and user preference vectors,\" Journal of Information Science and Engineering, vol. 35, no. 2, pp. 421–436, 2019 [Accessed: Apr. 2025] [14] M. Mintz, S. Bills, R. Snow, and D. Jurafsky, \"Distant supervision for relation extraction without labeled data,\" in Proc. ACL-IJCNLP, Singapore, 2009, pp. 1003–1011. [Accessed: Apr. 2025] [15] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, \"BERT: Pre-training of deep bidirectional transformers for language understanding,\" in Proc. NAACL-HLT, Minneapolis, MN, USA, 2019, pp. 4171–4186. [Accessed: Apr. 2025] [16] I. T. Jolliffe and J. Cadima, \"Principal component analysis: A review and recent developments,\" Philosophical Transactions of the Royal Society A, vol. 374, no. 2065, p. 20150202, 2016. [Accessed: Apr. 2025] [17] R. W. Sinnott, \"Virtues of the Haversine,\" Sky and Telescope, vol. 68, no. 2, p. 159, 1984. [Accessed: Apr. 2025] [18] N. Hug, \"Surprise: A Python library for recommender systems,\" Journal of Open Source Software, vol. 5, no. 52, p. 2174, 2020. [Accessed: Apr. 2025] [19] Z. Zhang, Q. Li, Z. Zeng, and H. Gao, \"User community discovery from multi-relational networks,\" in Proc. IEEE International Conference on Data Mining (ICDM), Brussels, Belgium, 2012, pp. 1078–1083. [Accessed: Apr. 2025] [20] L. Breiman, \"Random forests,\" Machine Learning, vol. 45, no. 1, pp. 5–32, 2001. [Accessed: Apr. 2025]

Copyright

Copyright © 2026 Dr. K. Chaitanya, M. Manaswi Priya, G. Praneetha, B. Pavan Kumar, N. Hari Krishna . This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.