Context

Friendship networks are very useful in recommender systems. For instance, by knowing homophily between two people, we can recommend them a product one of them has but not the other.

Objective

This project aims to develop an algorithm (and/or train a neural network) to predict if two individuals are friends or not using trajectory similarity and/or mobility habits. 

It is important for a dataset to have ground-truth for friendship information to achieve this objective. Therefore, we focus our research on the Breadcrumbs dataset (it is mandatory to work on this dataset). However, if students want to evaluate their solution on more datasets, they are encouraged to do so. Moreover, the Breadcrumbs dataset contain a “point-of-interest” table where each individual’s places are recorded. We plan a three-step plan for this project as follows:

Step 1: Implement the algorithm in [3].

Step 2: Evaluate the algorithm in [3].

Step 3: Improve the algorithm in [3] using other information from Breadcrumbs dataset such as answers from the “survey” table.

Step 4: Evaluate the improved algorithm in Step 3.

This subject as a Master’s Thesis requires you to read research papers on your project idea and take the lead in your Master’s Thesis.

If you want to learn more about the Breadcrumbs Dataset, we suggest you read the Breadcrumbs Paper [1] and Breadcrumbs Dataset Description [2].

Prerequisites

Students must be confident with their algorithms, mathematics, programming, data science, and machine learning skills. Preferred programming languages are Java and Python.

References

[1] Vaibhav Kulkarni, Arielle Moro, Bertil Chapuis, and Benoît Garbinato. 2017. Extracting Hotspots without A-priori by Enabling Signal Processing over Geospatial Data. In Proceedings of the 25th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (SIGSPATIAL ’17). Association for Computing Machinery, New York, NY, USA, Article 79, 1–4. DOI:https://doi.org/10.1145/3139958.3140002

[2] Breadcrumbs Dataset Description, https://github.com/doplab/breadcrumbsDB/blob/main/Breadcrumbs_Dataset_Description.pdf

[3] Quannan Li, Yu Zheng, Xing Xie, Yukun Chen, Wenyu Liu, and Wei-Ying Ma. 2008. Mining user similarity based on location history. In Proceedings of the 16th ACM SIGSPATIAL international conference on Advances in geographic information systems (GIS ’08). Association for Computing Machinery, New York, NY, USA, Article 34, 1–10. DOI:https://doi.org/10.1145/1463434.1463477

Contact

melike.gecer@unil.ch