Hotel Recommender System Menggunakan Metode Pendekatan Graph pada Dataset Trivago

Ricky Sunartio, Henry Novianus Palit, Andre Gunawan


Trivago is a global online application that provides service in searching for accomodations based on the user’s destination, and other specific filters such as price range, ratings, and available features. As an accommodation searching app, one of the major challenges that Trivago faces is providing a list of hotels that match the user’s preference, in order to increase the rate at which the user engages in a transaction. Trivago uses a recommender system that can deduce a user’s preference in a session, and displays a number of hotels that are considered suitable for that user. In an attempt to improve the quality of their system, Trivago works together with researchers from TU Wien, Politecnico di Milano, and Karlsruhe Institute of Technology, to conduct RecSys Challenge 2019 as an annual science competition under ACM Recommender Systems Conference.

This paper is going to be focused on the use of graph-based models in creating a recommender system using Trivago’s dataset provided in RecSys Challenge 2019. Graph-based models have been proven to be fairly effective in dealing with time series data, as shown in a research that studied online shoppers’ behavior using graphs [3]. Baumann’s research has shown an increase in accuracy by 5 – 10% when using graph features in predicting whether an online shopper would purchase a product as compared to using traditional features.

The graph-based model used in this research would be Markov Chain, a probabilistic graphical model. This research would test several Markov Chain models with varying length, depth, and order, measuring each model with a metric called Mean Reciprocal Rank (MRR) as per mentioned in the challenge. At the end of the research, it is concluded that the model that yields the highest MRR uses Markov Chain with length = 1, depth = 0, and order = 1.


Big Data; Recommender System; Trivago; Markov Chain

Full Text:



ACM. 2019. URI =

Bappalige, S. P. 2014, August 26. An introduction to Apache Hadoop for big data. URI =

Baumann, A., Haupt, J., Gebert, F., & Lessmann, S. 2018. Changing perspectives: Using graph metrics to predict purchase probabilities. Expert Systems With Applications 94, 137 - 148.

Brémaud, P. 2008. Markov Chains: Gibbs Fields, Monte Carlo Simulation, and Queues. Springer.

Chernev, A., Böckenholt, U., & Goodman, J. 2014. Choice overload: A conceptual review and meta-analysis. Journal of Consumer Psychology.

Kumar, R., Raghu, M., Sarlós, T., & Tomkins, A. 2017. Linear Additive Markov Process. The 26th International Conference on World Wide Web, (hal. 411-419).

Liu, S., & Wang, L. 2018. A self-adaptive point-of-interest recommendation algorithm based on a multi-order Markov model. Future Generation Computer Systems 89, 506 - 514.

Melnyk, S., Usatenko, O., & Yampol'skii, V. 2006. Memory functions of the additive Markov chains: applications to complex dynamic systems. Physica A: Statistical Mechanics and its Applications, 405-415.

Sejal, D., Ganeshsingh, T., Venugopal, K. R., Iyengar, S. S., & Patnaik, L. M. 2016. Image Recommendation Based on ANOVA Cosine Similarity. Procedia Computer Science 89, 562 - 567.

Trivago. 2019. Trivago. URI =


  • There are currently no refbacks.

Jurnal telah terindeks oleh :